1
00:00:00,040 --> 00:00:03,040
Mozilla's like never sat on the sidelines on stuff, like never.

2
00:00:03,330 --> 00:00:07,500
Going back 25 years, it's not our position to sit on the sidelines, like especially if...

3
00:00:08,920 --> 00:00:11,380
Hello, everybody, and welcome to Techlore Talks.

4
00:00:11,480 --> 00:00:15,280
Today, I am really excited to welcome back Mozilla, specifically Adam Fishman,

5
00:00:15,370 --> 00:00:18,960
who is their VP of product, to discuss his role in the company,

6
00:00:19,270 --> 00:00:23,940
the philosophy behind their recent AI tools at Mozilla, especially within the Firefox browser,

7
00:00:24,420 --> 00:00:27,520
the decision-making process behind them, how they work, the opt-out process,

8
00:00:27,660 --> 00:00:33,220
and why they chose to do it this way, and overall how Mozilla approaches AI as well as other features

9
00:00:33,500 --> 00:00:38,200
and what we can expect from them in the coming years and where their commitments lie for all of you to know about.

10
00:00:38,260 --> 00:00:43,200
It is no secret that AI is a polarizing technology, and when Mozilla started rolling things out in their browser,

11
00:00:43,720 --> 00:00:45,200
it's been a pretty polarizing thing.

12
00:00:45,480 --> 00:00:49,180
So I wanted to bring them on and ask them some questions that I had regarding it,

13
00:00:49,300 --> 00:00:51,920
so I hope that this interview can also answer some of yours.

14
00:00:52,260 --> 00:00:52,860
Let's get into it.

15
00:00:53,120 --> 00:00:56,840
Hello, everybody. I want to welcome Adam on the interview from Mozilla.

16
00:00:57,100 --> 00:01:01,300
if you don't mind sharing a little bit about what you do at Mozilla and a bit about yourself to get

17
00:01:01,440 --> 00:01:07,800
started. Yeah, sure. Thanks for having me, Henry. I am Mozilla's VP of product, which means that

18
00:01:08,480 --> 00:01:14,560
all of the product managers, the humans who are sort of responsible for figuring out what we

19
00:01:14,780 --> 00:01:20,360
should be building and where we should be taking the browser and features and functionality and

20
00:01:20,560 --> 00:01:25,059
stuff like that, like those are all part of my organization. We work with designers and engineers

21
00:01:25,060 --> 00:01:33,580
to build the browser. Got it. And then why does this role exist? I guess asking, not to call into

22
00:01:33,960 --> 00:01:40,340
question your value to the company, but can you just explain to me maybe a scenario in which

23
00:01:40,580 --> 00:01:47,340
having this role is necessary and what would happen if it wasn't there? Just operationally.

24
00:01:47,510 --> 00:01:52,339
For sure. Well, when companies are really small, you don't necessarily need this role. A lot of

25
00:01:52,360 --> 00:01:58,920
times in earlier stage startups, you have like the founder, the CEO is the head of the product

26
00:01:59,260 --> 00:02:03,980
organization. They basically say like, this is what we should build and why and everyone goes and

27
00:02:04,060 --> 00:02:08,899
builds it. But as companies get bigger, and you want to build and maintain more stuff, you have

28
00:02:08,899 --> 00:02:17,159
like a bigger and bigger surface area. And one of the roles of the product leader, myself, is

29
00:02:17,180 --> 00:02:22,180
basically to make sure that we're working on the most highly important stuff all the time.

30
00:02:22,570 --> 00:02:28,180
And so what happens is you have millions of ideas and I and the teams will bring those ideas

31
00:02:28,740 --> 00:02:34,360
through me or the sort of problems that they want to work on. And I help them sort of sift out like

32
00:02:34,700 --> 00:02:40,519
what's important to work on, what our users care about, what they don't. And so I kind of take like

33
00:02:40,560 --> 00:02:47,880
a huge like amount of stuff that we could go build and help us whittle it down to a smaller,

34
00:02:48,060 --> 00:02:53,160
more focused set of stuff so that we actually can build the right, the right things. Sometimes I get

35
00:02:53,160 --> 00:02:57,280
it wrong. Sometimes the team gets it wrong. That's one of the hard things about building products.

36
00:02:57,560 --> 00:03:03,380
But yeah, I work a lot with our, the executive team at Mozilla, my peers in engineering leadership.

37
00:03:04,040 --> 00:03:28,000
We're also kind of, again, as an organization gets a lot bigger, like we are the people who help figure out the way that work needs to move forward now and in different new and different ways. And so, yeah, how do we ship code, for example? How do we build products? Like, does that need to change? All that sort of stuff. That's kind of the role of like the leader of the product team.

38
00:03:28,400 --> 00:03:29,680
And the buck stops with me.

39
00:03:29,940 --> 00:03:33,980
I'm like the last and final decision maker on, should we launch a thing?

40
00:03:34,160 --> 00:03:35,200
Should we not launch a thing?

41
00:03:35,380 --> 00:03:36,180
Is it ready yet?

42
00:03:36,320 --> 00:03:36,900
Is it not ready?

43
00:03:37,440 --> 00:03:41,920
So I'm the ultimate accountable person when stuff goes out the door.

44
00:03:42,880 --> 00:03:43,120
Got it.

45
00:03:43,360 --> 00:03:52,960
So that's going to be relevant because I wanted to ask quite a bit about the AI stuff and see what the philosophy is there, but also some other features that have been dropped as well.

46
00:03:53,180 --> 00:04:01,960
And I wanted to see if maybe because as us, we might see a blog decision, but it'd be cool to get a little bit more of the internal thinking about what's going on back there.

47
00:04:02,340 --> 00:04:09,040
Before then, I also wanted to just ask kind of what's your history before Mozilla and like what brought you to Mozilla in the first place?

48
00:04:09,540 --> 00:04:13,500
I think what brought me to Mozilla is a lot of the same stuff that brings a lot of people to Mozilla.

49
00:04:13,980 --> 00:04:17,739
Like it is the last independent browser that exists.

50
00:04:18,000 --> 00:04:18,959
We're fighting the good fight.

51
00:04:19,079 --> 00:04:20,380
We have like incredible mission.

52
00:04:20,519 --> 00:04:24,820
I only work for companies that have a really great mission and a reason for existing.

53
00:04:25,370 --> 00:04:31,380
When you build products every day, it's really hard to do that work if you don't believe in,

54
00:04:31,650 --> 00:04:33,820
like very strongly in what the company is doing.

55
00:04:34,440 --> 00:04:35,840
So that's why I came to Mozilla.

56
00:04:35,950 --> 00:04:37,520
And also the people are really incredible.

57
00:04:37,800 --> 00:04:42,040
I had an opportunity to lead the new product development organization a year or two ago.

58
00:04:43,280 --> 00:04:47,160
And so like all of the things that aren't the web browser that we work on.

59
00:04:47,390 --> 00:04:49,340
And there's a million different products.

60
00:04:49,640 --> 00:04:50,340
It's really fun.

61
00:04:50,460 --> 00:04:52,180
It's very like startup-y.

62
00:04:52,500 --> 00:04:54,380
We're super scrappy and move really fast.

63
00:04:54,580 --> 00:04:56,500
And there's just some of the best people on the planet

64
00:04:56,760 --> 00:04:58,580
working across all of Mozilla.

65
00:04:58,960 --> 00:04:59,940
So it's really fun.

66
00:05:00,280 --> 00:05:00,760
It's great.

67
00:05:01,300 --> 00:05:02,260
We have a tough audience.

68
00:05:02,800 --> 00:05:03,580
You know, there's like,

69
00:05:04,140 --> 00:05:05,980
there's a lot of people that have really strong opinions

70
00:05:06,220 --> 00:05:07,200
on what we do.

71
00:05:07,340 --> 00:05:08,660
And they feel a lot of ownership,

72
00:05:08,960 --> 00:05:10,080
personal ownership in the browser.

73
00:05:10,540 --> 00:05:11,820
So that makes some days really hard.

74
00:05:12,040 --> 00:05:13,700
Like if we mess up or something like that.

75
00:05:14,120 --> 00:05:16,320
But, you know, generally we try to get it right.

76
00:05:16,700 --> 00:05:18,600
Even when we mess up, it's usually not intentional.

77
00:05:18,880 --> 00:05:20,500
And we usually fix it pretty quickly.

78
00:05:20,780 --> 00:05:22,000
So that's why I came to Mozilla.

79
00:05:22,180 --> 00:05:25,020
Before Mozilla, I worked at a bunch of different interesting companies.

80
00:05:25,080 --> 00:05:29,160
I worked for a company called Imperfect Foods, which ships ugly fruits and vegetables to people.

81
00:05:29,940 --> 00:05:30,780
I think I tried them once.

82
00:05:31,080 --> 00:05:31,240
Yeah.

83
00:05:32,000 --> 00:05:34,460
I worked for Lyft.

84
00:05:34,940 --> 00:05:39,880
I worked for Patreon, where you are a creator on the platform for a little bit longer.

85
00:05:40,360 --> 00:05:40,420
Yeah.

86
00:05:41,000 --> 00:05:46,579
And always in like building products and always in mission-driven and mission-oriented companies

87
00:05:46,600 --> 00:05:53,120
like Imperfect Foods, Patreon, like even Lyft, all very mission driven businesses. And that's

88
00:05:53,120 --> 00:05:57,480
been super important for me for my my whole career, just makes it worth getting out of bed in the

89
00:05:57,620 --> 00:06:02,500
morning to go to work. So that's what I do. Very nice. And then I think a really good starting

90
00:06:02,660 --> 00:06:08,300
point here actually isn't I thought of it as you were talking, pocket, right? Like pocket, I think

91
00:06:08,300 --> 00:06:13,819
is maybe a good starting point to kind of discuss things that, you know, maybe made some some of the

92
00:06:13,840 --> 00:06:19,500
community upset. I mean, I was one of the people who was quite anti-Pocket. I never used it. One

93
00:06:19,500 --> 00:06:25,300
of the first things I did when I installed Firefox is disable Pocket. And I kind of would always

94
00:06:25,860 --> 00:06:32,240
humorously say, I don't know who's using this. But after you removed Pocket, they all showed up.

95
00:06:32,320 --> 00:06:38,379
They all spawned. But what was the behind the scenes look? I don't know if you were involved

96
00:06:38,400 --> 00:06:44,240
in Mozilla when Pocket was dropped. But if you have any context there, it's just kind of a starting

97
00:06:44,420 --> 00:06:49,600
point of something that maybe people are a little more familiar with. Yeah, I was involved in not

98
00:06:49,780 --> 00:06:55,460
necessarily the decision to sunset or close down Pocket, but I was definitely involved in the

99
00:06:55,740 --> 00:07:00,620
discussion around what is this thing? What are we doing with it? I think the reality is like Pocket

100
00:07:00,680 --> 00:07:08,940
was a company that we acquired a long time ago when news like aggregation and sort of like save

101
00:07:08,940 --> 00:07:15,100
it for later tools were pretty popular growing like gangbusters. And it made a lot of sense for

102
00:07:15,580 --> 00:07:20,940
the web and how people were navigating the web. And then what we found was over time, like it

103
00:07:21,140 --> 00:07:26,999
required a lot of energy and effort to maintain like a lot, an outsized amount of energy and effort

104
00:07:27,060 --> 00:07:32,480
the amount of people that were using it. And at the end of the day, we were looking at the people

105
00:07:32,680 --> 00:07:37,960
working on it and what it was doing to advance the efforts of the browser and Mozilla's mission.

106
00:07:38,660 --> 00:07:43,240
And we just sort of came to this conclusion that like, could we take these people that are working

107
00:07:43,380 --> 00:07:49,320
on Pocket and have them work on other stuff that is more meaningful, that is more of what our users

108
00:07:49,560 --> 00:07:53,939
want. And even though there was a passionate audience of Pocket users, that passionate

109
00:07:53,960 --> 00:07:59,960
audience wasn't growing and it was declining pretty steadily over time. And there were a bunch

110
00:08:00,100 --> 00:08:09,040
of alternatives to Pocket. And we just were like, do we, do we, should we be in this, in this world

111
00:08:09,280 --> 00:08:14,560
anymore? As much as we love it, we were like, what if the smart people working on this could do a

112
00:08:14,560 --> 00:08:19,819
bunch of other really great things for the browser and what would happen? And obviously we knew that

113
00:08:19,840 --> 00:08:25,740
would be a bummer for the people that use and dependent on Pocket. But we tried to provide a

114
00:08:25,770 --> 00:08:29,740
smooth transition for folks to find other things that they could use. And there's other companies

115
00:08:29,960 --> 00:08:34,520
that we sort of endorsed and things like that. But yeah, at the end of the day, we were just like,

116
00:08:34,820 --> 00:08:39,479
gosh, we have so many more things that we want to do. And we are sinking a lot of time and energy

117
00:08:40,060 --> 00:08:46,719
into this particular product. And even though a number of people like it, we know we could build

118
00:08:46,740 --> 00:08:52,940
a lot better things for even more people if we put that time and attention elsewhere.

119
00:08:53,460 --> 00:08:57,540
It's just one of those tough decisions that like winding it down was winding down any product

120
00:08:57,740 --> 00:09:01,800
is a really hard decision, like especially if a lot of people are using it.

121
00:09:02,200 --> 00:09:06,200
It's it's one of the hardest things that a product manager or product leader has to do.

122
00:09:07,480 --> 00:09:11,700
So, you know, you don't take it lightly, but we were able to like keep all the people that

123
00:09:11,740 --> 00:09:14,120
were working on pocket and have them work on other great stuff.

124
00:09:14,300 --> 00:09:20,540
So like some of them are now basically building parts of the desktop version of Firefox or

125
00:09:20,540 --> 00:09:24,000
the mobile version or like on the content team.

126
00:09:24,300 --> 00:09:29,440
And so like all of that stuff, they've gone on to greater homes at Mozilla and at Firefox

127
00:09:29,600 --> 00:09:34,380
to have like bigger impact on other stuff that feels more relevant to today.

128
00:09:35,120 --> 00:09:36,080
Yeah, very nice.

129
00:09:36,240 --> 00:09:37,680
Yeah, I know it's always challenging.

130
00:09:37,800 --> 00:09:42,820
And so I think just hearing like what happens behind the scenes is quite valuable.

131
00:09:43,260 --> 00:09:47,140
And I guess, you know, so that's something that was removed in the past.

132
00:09:47,310 --> 00:09:52,860
But I feel like I also now like to see kind of the internal processing of something that was added, especially AI.

133
00:09:53,430 --> 00:09:54,600
Very polarizing topic.

134
00:09:55,060 --> 00:09:55,440
For sure.

135
00:09:55,660 --> 00:09:57,480
You guys really stuck your necks out.

136
00:09:57,850 --> 00:10:02,920
As you said, the community is quite precise on what they want.

137
00:10:02,990 --> 00:10:09,120
And I feel like you did the one thing that polarizes any community that already is polarized on this topic.

138
00:10:09,660 --> 00:10:09,780
Yeah.

139
00:10:11,140 --> 00:10:18,800
So I want to start with just your philosophy, not I guess maybe you personally, but also just as Mozilla, like what is your guys' philosophy on AI?

140
00:10:18,890 --> 00:10:22,240
I feel like now in 2026, things are settling a bit down.

141
00:10:22,370 --> 00:10:31,180
I feel like up until even early 2025, there were still these, everyone just had these bright rose colored glasses on.

142
00:10:31,660 --> 00:10:32,800
This is going to fix everything.

143
00:10:33,110 --> 00:10:34,460
The hype was just very high.

144
00:10:34,880 --> 00:10:37,100
And I feel like the hype has died down quite a lot.

145
00:10:37,130 --> 00:10:38,480
But these companies haven't gone anywhere.

146
00:10:38,750 --> 00:10:39,620
AI hasn't gone anywhere.

147
00:10:40,500 --> 00:10:43,540
So what is your philosophy on AI, just to kind of open up this discussion?

148
00:10:44,160 --> 00:10:48,480
So I have my personal philosophy, which is fairly well aligned with Mozilla's philosophy.

149
00:10:48,630 --> 00:10:51,060
So I'll kind of speak about both of these things at the same time.

150
00:10:51,400 --> 00:10:56,060
I think, and especially for our audience who thinks about this a lot, I think a lot of

151
00:10:56,220 --> 00:11:03,760
people have every right to be skeptical of AI that's kind of like gimmicky and forced.

152
00:11:04,260 --> 00:11:08,959
And there is a whole world of products being created where people are just like, well, if

153
00:11:08,980 --> 00:11:10,480
We just sprinkle some AI on it.

154
00:11:10,640 --> 00:11:12,060
That's that that's that'll work.

155
00:11:12,260 --> 00:11:20,080
And a lot of that is coming from like outside pressure, you know, boardrooms, investors, public markets, like things like that.

156
00:11:20,230 --> 00:11:25,300
And so like I don't and Mozilla doesn't subscribe to that theory of AI.

157
00:11:25,840 --> 00:11:27,200
Like it's a technology.

158
00:11:28,100 --> 00:11:28,600
It's interesting.

159
00:11:28,950 --> 00:11:30,940
It can do some really interesting things.

160
00:11:31,480 --> 00:11:42,900
I think regardless of how you feel about AI, I think one thing that's pretty clear is we're sort of working, moving towards this new era of how people and computers interact with each other.

161
00:11:43,350 --> 00:11:50,780
And AI is going to have a really big impact on shaping access and the use of the Internet and information.

162
00:11:51,560 --> 00:11:54,960
And that has always been Mozilla's focus.

163
00:11:55,760 --> 00:12:02,620
And so, you know, we could put our head in the sand and sit on the sidelines and be like,

164
00:12:02,800 --> 00:12:04,800
no, we're not going to pay any attention to us.

165
00:12:05,290 --> 00:12:10,760
And then the problem is Google and Microsoft and OpenAI, they're going to make all those

166
00:12:10,940 --> 00:12:11,200
decisions.

167
00:12:12,140 --> 00:12:13,440
And that's not good.

168
00:12:13,920 --> 00:12:18,280
Mozilla's like never sat on the sidelines on stuff, like never going back 25 years.

169
00:12:18,760 --> 00:12:23,139
It's not our it's not our position to sit on the sidelines, like especially if big things

170
00:12:23,160 --> 00:12:26,020
are shaping the future of how people use the web.

171
00:12:26,370 --> 00:12:29,800
And so like our philosophy is we're going to play the role here

172
00:12:29,800 --> 00:12:31,800
that we've played for 25 years,

173
00:12:31,990 --> 00:12:35,440
which is we're going to be the stewards of the right way to do things.

174
00:12:35,770 --> 00:12:37,280
And we're going to be in the conversation

175
00:12:37,450 --> 00:12:38,920
and we're going to work on products

176
00:12:39,330 --> 00:12:43,600
so that we can have a say in how AI works, evolves,

177
00:12:43,780 --> 00:12:46,600
is actually useful to people, is trustworthy,

178
00:12:47,540 --> 00:12:52,039
and puts the users of the browser, the users of the AI in control

179
00:12:52,060 --> 00:12:53,820
and doesn't like thrust it upon them.

180
00:12:54,220 --> 00:12:56,160
Some people are not going to be happy by that stance.

181
00:12:56,400 --> 00:12:56,840
They're going to be like,

182
00:12:56,920 --> 00:12:59,320
you shouldn't even touch this with a 10 foot pole.

183
00:12:59,940 --> 00:13:02,080
And I just think we feel differently about it.

184
00:13:02,640 --> 00:13:04,620
You can, there is a middle ground here

185
00:13:04,700 --> 00:13:06,720
where you can have the values and the views

186
00:13:06,880 --> 00:13:09,660
that Mozilla has and believe that it's important

187
00:13:09,700 --> 00:13:11,140
for us to be involved in this

188
00:13:11,480 --> 00:13:13,800
or else somebody else is going to write all the rules

189
00:13:14,619 --> 00:13:16,900
and people are going to be way worse off as a result.

190
00:13:17,260 --> 00:13:20,080
So we choose to participate instead of sitting it out.

191
00:13:20,840 --> 00:13:25,700
Yeah, I think this is, as an outsider looking in,

192
00:13:27,060 --> 00:13:30,260
it calls into question the nuance of what you all are.

193
00:13:30,540 --> 00:13:34,040
Because if I see you all as just a browser, just Firefox,

194
00:13:35,100 --> 00:13:41,640
it is quite hard to justify the AI position in some ways, I think, just on its face.

195
00:13:42,440 --> 00:13:47,120
But when I hear what you say, I hear Mozilla sees itself as,

196
00:13:47,560 --> 00:13:50,180
no matter what the feature they release, it doesn't have to be AI,

197
00:13:50,380 --> 00:13:53,720
could be some made up thing, you know, like we can make something up, let's say machine learning,

198
00:13:54,540 --> 00:13:58,820
which is right now, maybe that was like the pre phrasing similar feature sets of AI,

199
00:13:59,280 --> 00:14:03,660
maybe it's, I don't know, googly bears, you know, like some new technology, it doesn't exist.

200
00:14:04,460 --> 00:14:09,660
And these big tech companies are starting to implement it. Independent of the Firefox browser,

201
00:14:09,840 --> 00:14:14,220
Mozilla would try to still be at that table and do what it takes to be at the table to still

202
00:14:15,100 --> 00:14:19,800
try to represent the values of that community. Do you think it's a fair assessment?

203
00:14:20,000 --> 00:14:40,140
Yeah, I think so. Because we do. I mean, there's Mozilla, the foundation, right, which is all the advocacy work and all the pushing that we do with governments and regulators and all of that sort of thing to advocate for people, right, for choice, for like anti monopoly behavior and things like that.

204
00:14:40,560 --> 00:14:45,260
Like that is the world that we're going to operate in regardless of the Firefox browser.

205
00:14:45,420 --> 00:14:47,740
It's just like how we see the world.

206
00:14:48,220 --> 00:14:50,580
And so, yeah, I think you're right on that front.

207
00:14:50,940 --> 00:14:52,580
So do you see this AI?

208
00:14:53,440 --> 00:14:55,820
Because people are seeing the AI as a Firefox feature.

209
00:14:57,180 --> 00:15:01,120
So how much of the AI rollouts do you feel like is a feature?

210
00:15:01,620 --> 00:15:06,340
Like that's designed to help users and be in, like, let's say you didn't have any advocacy.

211
00:15:06,410 --> 00:15:09,180
You weren't going to be showing up at any table with big tech companies.

212
00:15:09,320 --> 00:15:10,700
you were just managing a browser.

213
00:15:11,820 --> 00:15:13,300
So how much of this AI do you see

214
00:15:13,400 --> 00:15:15,480
as genuinely useful for people

215
00:15:15,880 --> 00:15:16,900
and like fits into their workflow

216
00:15:17,140 --> 00:15:19,040
versus more of that advocacy role?

217
00:15:19,380 --> 00:15:22,060
Like this is something that we need to figure out

218
00:15:22,380 --> 00:15:23,200
what this looks like,

219
00:15:23,300 --> 00:15:24,500
like what does AI look like

220
00:15:24,640 --> 00:15:26,160
when it's actually done in a proper way?

221
00:15:26,780 --> 00:15:28,200
And you're kind of building the infrastructure

222
00:15:28,420 --> 00:15:31,480
for better companies, organizations, individuals

223
00:15:31,960 --> 00:15:33,280
to have a better avenue.

224
00:15:33,380 --> 00:15:35,660
Like they don't have to go through some Chrome API

225
00:15:35,720 --> 00:15:38,100
that only works with their, like, you know.

226
00:15:38,420 --> 00:15:43,660
Yeah. Well, the answer is kind of both. We kind of do both at the same time, right? And it's hard

227
00:15:43,710 --> 00:15:48,380
because it means you're doing a lot. But I'd say a couple of things about that. One is, you know,

228
00:15:48,620 --> 00:15:53,740
we talk to people, you probably have seen our Reddit presence. You probably have seen people

229
00:15:53,840 --> 00:15:59,700
to Reddit on my DNS. So no, I did it. I did it for I can't take right. Yeah, well, I was gonna say

230
00:15:59,900 --> 00:16:06,360
for your own mental health reasons. Sorry to cut you off. No, no, no, it's totally fine. But like,

231
00:16:06,980 --> 00:16:10,620
there's an active and vibrant Mozilla community in a lot of places, right?

232
00:16:10,800 --> 00:16:12,920
Social media, Reddit, like our own connect,

233
00:16:13,580 --> 00:16:16,540
like we get inbound all the time from, from folks.

234
00:16:17,100 --> 00:16:18,340
And, you know,

235
00:16:18,400 --> 00:16:23,640
we asked a lot of people at all ends of the spectrum of like comfort with AI,

236
00:16:23,780 --> 00:16:28,120
like, did they want to, were they curious about it? Did they like,

237
00:16:28,220 --> 00:16:30,580
what do they want to see in the browser?

238
00:16:31,140 --> 00:16:35,140
And the majority of people said like, I'm curious about this.

239
00:16:35,700 --> 00:16:41,600
I'm curious what Mozilla's take is on how it can make the browser better or our experience better.

240
00:16:42,040 --> 00:16:45,720
And so like, yeah, I want you to do stuff here.

241
00:16:46,260 --> 00:16:50,080
And so we said, OK, that gives us some good license to experiment.

242
00:16:50,520 --> 00:16:52,760
Like, let's learn how this can be helpful.

243
00:16:53,020 --> 00:16:55,480
Let's learn what the Mozilla approach to AI is.

244
00:16:55,680 --> 00:17:00,780
And if we don't do that, we're actually ignoring a huge, like the majority part of the users of the browser.

245
00:17:01,260 --> 00:17:03,340
So like that was like one very strong signal.

246
00:17:03,760 --> 00:17:08,520
In terms of like what is useful with AI, I think a lot of people think about AI and they're like,

247
00:17:08,569 --> 00:17:13,800
they think of just the generative AI that I know today, the like slop that's filling the internet

248
00:17:13,970 --> 00:17:20,199
with stuff, the, you know, fake images and like all that's hallucinations and like all of that

249
00:17:20,230 --> 00:17:25,380
sort of stuff. But there's a lot more to it that I think people don't realize is like, oh, actually

250
00:17:25,480 --> 00:17:31,479
that's AI as a technology under the hood doing stuff. Accessibility is an example. And you

251
00:17:31,500 --> 00:17:34,680
You probably know this Mozilla is a company that cares a great deal about accessibility

252
00:17:35,000 --> 00:17:39,880
because the more people who can access the internet and content and the more it's sort

253
00:17:39,880 --> 00:17:42,100
of democratized, the better, right?

254
00:17:42,120 --> 00:17:43,200
More people online, better.

255
00:17:43,580 --> 00:17:50,320
As you navigate the web, there are an infinite number of images, for example, that don't have

256
00:17:50,340 --> 00:17:51,820
any alt text associated with them.

257
00:17:52,400 --> 00:17:56,160
So we have an on-device AI model in Firefox.

258
00:17:56,400 --> 00:17:57,220
So it's on your device.

259
00:17:57,860 --> 00:18:00,620
So that means it's not like hitting a cloud server.

260
00:18:00,760 --> 00:18:02,160
It's not sending data anywhere else.

261
00:18:02,290 --> 00:18:03,560
It's literally doing this on your device.

262
00:18:04,400 --> 00:18:06,680
That as it comes across one of those images,

263
00:18:06,790 --> 00:18:08,320
the browser's rendering the page,

264
00:18:08,900 --> 00:18:11,140
it will check for alt text.

265
00:18:11,720 --> 00:18:13,280
It will analyze that image

266
00:18:13,390 --> 00:18:18,000
and it will actually output a description of the image as alt text.

267
00:18:18,680 --> 00:18:21,340
So if you're like a very anti-AI person

268
00:18:21,880 --> 00:18:23,380
and you're like no AI anywhere,

269
00:18:24,080 --> 00:18:25,540
what if you were also blind

270
00:18:26,220 --> 00:18:27,500
and you needed that information

271
00:18:27,700 --> 00:18:30,540
because most people don't add their own alt text to images?

272
00:18:30,640 --> 00:18:37,100
on the web. Like that is a genuinely useful feature to a large swath of the population.

273
00:18:37,140 --> 00:18:41,200
I think we estimate that something like one in five or one in four people who use the web has

274
00:18:41,260 --> 00:18:45,800
some sort of disability. Now they're not all blind, of course, but like imagine the population that is,

275
00:18:46,000 --> 00:18:50,360
that is a really useful feature. And there's no way that we can do that at scale. Like no human

276
00:18:50,580 --> 00:18:55,300
can really do that at scale. And so like, that's a very specific, narrow, like feel good example.

277
00:18:55,760 --> 00:19:00,620
But the other side of this is we have this new product that's coming out. It probably around

278
00:19:00,640 --> 00:19:04,300
the time that this show is going to release, we're going to have this product that's coming out in

279
00:19:04,420 --> 00:19:10,140
beta called Smart Window. And we've talked about this a lot. Basically, first, you can choose to

280
00:19:10,320 --> 00:19:14,560
use it or not. So one of our philosophies about AI, we could talk about AI controls too, if you want

281
00:19:14,560 --> 00:19:19,720
to talk about that in a sec. But like, you can turn it all off if you want. And that is, again,

282
00:19:20,020 --> 00:19:25,020
a feature that I have not seen any other browser do, is the ability to have a toggle that says,

283
00:19:25,140 --> 00:19:29,160
like, get all this stuff out of my face, delete all the models that are locally on my machine,

284
00:19:29,380 --> 00:19:34,080
like gone. I don't want to be aware of it. I don't want to be exposed to it. I don't even want to see

285
00:19:34,080 --> 00:19:38,640
a marketing message that says you should use it. Like you can turn it all off. And then you can

286
00:19:38,700 --> 00:19:44,320
like selectively reactivate stuff if you want. That's fine. But anyway, so smart window, it

287
00:19:44,720 --> 00:19:51,720
basically is another browsing experience. Again, you have to go into it yourself. We're not like

288
00:19:51,860 --> 00:19:57,700
saying you must use this. It's almost like an incognito window or a private window. You know,

289
00:19:57,840 --> 00:20:02,040
we have private window, we have a regular window, classic window, and then we'll have a smart window.

290
00:20:02,240 --> 00:20:06,460
So you can go into this and you can say, hey, because I'm going into this, you can say, I'm

291
00:20:06,920 --> 00:20:14,680
interested in AI, right? And it will, as you're navigating the web, develop a memory of your

292
00:20:14,900 --> 00:20:21,760
behaviors on the web. Again, locally stored, like on your device. So it's, it knows like some things

293
00:20:21,880 --> 00:20:26,780
you're interested in. It knows that you are a vegetarian based on things that you're doing.

294
00:20:26,900 --> 00:20:29,100
You can also see and remove specific memories.

295
00:20:29,480 --> 00:20:31,660
So you know everything that it knows about you.

296
00:20:32,160 --> 00:20:33,600
And you can say like, I don't like that.

297
00:20:33,800 --> 00:20:34,280
Get rid of it.

298
00:20:34,280 --> 00:20:35,400
I don't want you to know that.

299
00:20:35,700 --> 00:20:36,140
Take that out.

300
00:20:36,380 --> 00:20:39,380
Is an example of like how this is genuinely useful to somebody.

301
00:20:40,060 --> 00:20:45,240
During our piloting of this, we have somebody at Mozilla who is building this product.

302
00:20:45,960 --> 00:20:48,020
And this person is really into heavy metal.

303
00:20:48,760 --> 00:20:48,920
Okay.

304
00:20:49,480 --> 00:20:54,420
Not my particular genre of music, but you know, no hate for anyone that loves heavy metal.

305
00:20:54,620 --> 00:20:55,520
They also play guitar.

306
00:20:56,400 --> 00:21:00,360
And so they were looking at two different guitars and they were doing like comparison shopping.

307
00:21:01,000 --> 00:21:02,420
And they were like, which one of these should I get?

308
00:21:03,419 --> 00:21:08,080
So the smart window, there's an assistant that kind of follows you around.

309
00:21:08,580 --> 00:21:15,900
And it basically said, like, you should think about this guitar between these two because it is better for heavy metal.

310
00:21:16,780 --> 00:21:19,060
And it was like, whoa, why is it better for heavy metal?

311
00:21:19,180 --> 00:21:45,560
It's like, well, it doesn't say that in the description of the guitar, but it knows that the guitar has a longer string, which you can use for tuning a guitar lower and making it heavier and like, like making it sound heavier and stuff like that. And so like, this is an example where, okay, it knows this is a subtle thing about you that you like heavy metal. And now you're doing something on the internet, you're shopping for a guitar, you're browsing, and it's making a helpful recommendation for somebody.

312
00:21:45,620 --> 00:21:51,880
So that's like not the accessibility example, but it's an example of just like how it's how it's useful.

313
00:21:52,620 --> 00:21:55,060
Would the person been able to figure this out eventually on their own?

314
00:21:55,610 --> 00:22:03,780
Probably, but would have required a lot more research, like would have required like knowing specifically about the string length and like stuff like that.

315
00:22:03,880 --> 00:22:07,920
And just it, you know, the AI can help you with that with that sort of stuff.

316
00:22:08,460 --> 00:22:12,900
And again, it's really only for people who absolutely want to use it because they don't have to otherwise.

317
00:22:13,580 --> 00:22:24,880
So but both of those things seem pretty like genuinely useful to me as someone who like buys things online, you know, as someone who knows a lot of people who have disabilities that are trying to use the web and things like that.

318
00:22:24,920 --> 00:22:27,000
And those are two very different use cases of AI.

319
00:22:27,400 --> 00:22:29,840
You can actually turn both of those off if you want.

320
00:22:30,040 --> 00:22:37,920
But we found that a lot of people would turn off, even if they were turning off the AI controls, they would then selectively turn back on something.

321
00:22:38,120 --> 00:22:40,920
They're like, oh, actually, this is really useful to me, but I'm gonna leave this on.

322
00:22:41,500 --> 00:22:47,160
And so I found that really interesting in talking to people about how they use the features, too.

323
00:22:48,340 --> 00:22:50,820
Very nice. Yeah. And just, I guess, a quick comment.

324
00:22:51,180 --> 00:22:58,720
It's interesting because I feel like maybe until offline AI models, and maybe this is something I need to evolve my own views on,

325
00:22:58,780 --> 00:23:05,780
because I hold this view that personalization is a bad word, actually,

326
00:23:06,120 --> 00:23:12,700
because the way personalization has been done until now has been cloud,

327
00:23:13,680 --> 00:23:14,520
collecting lots of data,

328
00:23:15,080 --> 00:23:18,960
and then inferring your preferences based on that from big tech companies.

329
00:23:19,480 --> 00:23:21,920
And so anytime a service says personalization,

330
00:23:22,360 --> 00:23:23,940
it's typically for me a big red flag.

331
00:23:25,600 --> 00:23:27,620
That's still the case, like 90% of the times.

332
00:23:27,780 --> 00:23:43,940
But I think this offline AI world is now enabling offline tools to be able to do this personalization without needing the cloud, which hasn't, I mean, it could have been possible before, but we're seeing it a lot more often now where it's like, oh, we now have an offline way of analyzing things, et cetera.

333
00:23:44,480 --> 00:23:45,980
Just a comment, I guess.

334
00:23:46,260 --> 00:23:48,860
I wanted to just get a tiny bit technical here.

335
00:23:49,200 --> 00:23:49,320
Sure.

336
00:23:49,420 --> 00:23:59,140
I don't know if you guys have a very, you know, you talk about this idea of showing up at the table and making sure that you're representing, you know, a better way of doing things.

337
00:23:59,900 --> 00:24:13,680
Can you point to like a specific thing that you've seen other big tech companies propose that concerns you that you think people should know about that you at Mozilla are actively doing differently and how that could play into things?

338
00:24:13,800 --> 00:24:20,860
I feel like we see this in the in like the web engine space a lot where Google wants to do something with Chromium and then you guys like, no, we need to do it this way.

339
00:24:21,000 --> 00:24:22,940
So I don't know. Yeah. If you have something like that.

340
00:24:23,380 --> 00:24:31,100
I don't know that I have a specific example related to like the browser engine or something, something like super highly technical like that.

341
00:24:31,100 --> 00:24:33,700
I mean, I know that we are doing that stuff all the time.

342
00:24:33,880 --> 00:24:35,100
In the AI side of things.

343
00:24:35,360 --> 00:24:37,160
In the AI side of things.

344
00:24:37,560 --> 00:24:45,920
You know, we haven't yet seen a ton of legislation around AI usage, but we're involved in all

345
00:24:45,990 --> 00:24:47,840
of those conversations right now.

346
00:24:48,230 --> 00:24:54,860
I would say where we differ from big tech in terms of like big tech does it one way and

347
00:24:54,860 --> 00:24:58,180
we do it a different way are a couple in a couple dimensions.

348
00:24:59,040 --> 00:25:02,820
One is we've been talking about this idea of choice.

349
00:25:03,680 --> 00:25:06,000
And this is choice comes in a lot of shapes and sizes.

350
00:25:06,160 --> 00:25:08,900
So like take the Chrome browser, for example.

351
00:25:09,280 --> 00:25:10,720
They have an AI in Chrome.

352
00:25:11,740 --> 00:25:13,280
That AI is Google's AI.

353
00:25:13,760 --> 00:25:14,560
They own that AI.

354
00:25:15,500 --> 00:25:20,880
They want that AI to have as much information about you as possible for the reasons that you said.

355
00:25:21,100 --> 00:25:22,800
It makes better ad targeting.

356
00:25:23,140 --> 00:25:27,020
Like at the end of the day, we do not own an LLM.

357
00:25:27,540 --> 00:25:28,340
We're not training them.

358
00:25:28,920 --> 00:25:30,980
We're not in the business of doing that.

359
00:25:31,130 --> 00:25:34,700
That is one, reserved for people with way deeper pockets than us.

360
00:25:35,930 --> 00:25:37,980
And two, like, you don't need it.

361
00:25:38,280 --> 00:25:41,140
There's like a million awesome open source AI models now.

362
00:25:41,580 --> 00:25:49,040
So we are giving people options and choice on like, which AI model would you like to use in Smart Window?

363
00:25:49,540 --> 00:25:53,760
And we're telling you like, well, this one's faster and this one's better at this and this one's better at this.

364
00:25:54,080 --> 00:26:02,940
And then the other thing is we're also allowing people to bring their own model, which is a slightly more technical feature.

365
00:26:03,620 --> 00:26:06,940
Eventually, we're going to make it super easy for anyone to do it.

366
00:26:07,240 --> 00:26:21,160
But right now, you can actually go into the smart window when it's live, and you can say, using the open router AI protocol, you can say, I want this model that lives here.

367
00:26:21,480 --> 00:26:24,540
And you can even have it be a local one and point it at that.

368
00:26:24,880 --> 00:26:25,580
And it just works.

369
00:26:26,280 --> 00:26:30,020
So you guys use right now, it's only open router through like an API key.

370
00:26:30,520 --> 00:26:31,940
Do you do you like Olama or anything?

371
00:26:32,840 --> 00:26:38,040
Yes, I think we're looking at a lot of different ways that we can work like different partners

372
00:26:38,160 --> 00:26:39,220
that we can work with here.

373
00:26:40,180 --> 00:26:44,640
But open routers like the sort of the most common protocols are the ones that we're trying

374
00:26:44,640 --> 00:26:47,980
to use in the beginning because it makes it most widely accessible to people.

375
00:26:48,600 --> 00:26:51,900
But like, that's an example, like nobody else is going to do that.

376
00:26:52,240 --> 00:26:55,560
Like chat GPT's browser is not going to use any other model.

377
00:26:55,640 --> 00:26:57,780
It's going to use, it's going to use chat GPT.

378
00:26:58,120 --> 00:26:58,680
Like, come on.

379
00:26:59,200 --> 00:27:02,180
I should say open AI's browser, Microsoft, same deal.

380
00:27:02,640 --> 00:27:04,740
Like we are trying to give people that choice.

381
00:27:05,320 --> 00:27:07,240
And then the other thing is like the idea of turning it off.

382
00:27:07,720 --> 00:27:10,480
Like nobody else is really letting people turn it off.

383
00:27:10,600 --> 00:27:12,060
Certain none of the big companies.

384
00:27:12,680 --> 00:27:18,300
And like, we know that there are, there's a world where if you turn off AI in the, in

385
00:27:18,320 --> 00:27:21,220
it might actually make your browser experience

386
00:27:22,280 --> 00:27:24,560
not as great as somebody who's enhancing it

387
00:27:24,680 --> 00:27:26,700
with AI stuff like translation

388
00:27:27,080 --> 00:27:28,620
and sort of all of those things.

389
00:27:29,200 --> 00:27:30,480
But like, that's okay.

390
00:27:30,940 --> 00:27:31,620
That's people's choice.

391
00:27:32,280 --> 00:27:35,500
And so I think this idea of like choice and agency

392
00:27:36,300 --> 00:27:37,440
and then also transparency,

393
00:27:37,840 --> 00:27:39,080
like who else is showing people

394
00:27:39,260 --> 00:27:43,680
like what memories the model has about them?

395
00:27:43,800 --> 00:27:45,940
Like, is Google gonna tell you all the data

396
00:27:46,080 --> 00:27:46,940
that it has about you?

397
00:27:47,020 --> 00:27:47,920
Like, probably not.

398
00:27:48,220 --> 00:27:54,500
that's a scary proposition, you know? Um, and we show everybody that in like full transparency,

399
00:27:54,920 --> 00:28:00,080
like pull the curtain back. Here's everything it knows about you and delete the stuff that you don't

400
00:28:00,160 --> 00:28:03,500
want it to know about you or turn it off entirely. So you, you don't want it to collect any

401
00:28:03,660 --> 00:28:08,020
information. That's fine. Um, it's quite similar. Actually, I think Claude might be the only one

402
00:28:08,020 --> 00:28:12,400
that seems to do that because Claude has memory and also Claude doesn't have a browser. They just

403
00:28:12,500 --> 00:28:17,560
have their extension, which I think only works in Chrome so far. Only works in Chrome. Um,

404
00:28:17,700 --> 00:28:19,440
But it would be cool, I think.

405
00:28:20,020 --> 00:28:23,700
You know, we seem to be a bit more values aligned

406
00:28:23,900 --> 00:28:24,980
with the team at Anthropic.

407
00:28:25,200 --> 00:28:26,920
We've actually done some cool partnership stuff.

408
00:28:26,990 --> 00:28:28,800
I saw the security vulnerabilities thing.

409
00:28:29,030 --> 00:28:30,980
Oh, man, that was amazing.

410
00:28:31,480 --> 00:28:32,580
Yeah, that's actually,

411
00:28:32,890 --> 00:28:34,820
I would love to tell that story a little bit,

412
00:28:35,040 --> 00:28:37,280
but we can do that later if you want.

413
00:28:37,440 --> 00:28:37,800
That's fine.

414
00:28:37,920 --> 00:28:38,560
I mean, we're here.

415
00:28:39,480 --> 00:28:41,600
I have my list of things to touch on.

416
00:28:41,810 --> 00:28:42,260
Okay, cool.

417
00:28:42,780 --> 00:28:46,480
So basically, so you mentioned security vulnerabilities.

418
00:28:46,880 --> 00:28:55,060
So basically what happened was the folks at Anthropic who build Claude, the various Claude models, they have a like a red teaming security team.

419
00:28:55,740 --> 00:29:05,440
And they were building and training models to find security vulnerabilities in products, which seems like could be a good thing, could be a terrifying thing.

420
00:29:05,550 --> 00:29:08,320
We had previously tried this with AI, like a year ago.

421
00:29:08,390 --> 00:29:10,200
We were like, we're curious, is this work?

422
00:29:10,620 --> 00:29:11,160
And it didn't work.

423
00:29:11,660 --> 00:29:14,460
Like it was, the models were not very good at the time for this sort of stuff.

424
00:29:15,020 --> 00:29:17,420
And then some new models came out.

425
00:29:17,980 --> 00:29:25,300
And what happened was the Anthropic team started testing a bunch of the JavaScript engine in Firefox.

426
00:29:25,860 --> 00:29:31,180
And the reason that they did it is they're like, well, this one's going to be a tough nut to crack because it's open source.

427
00:29:31,310 --> 00:29:33,720
So the community has been pounding on this for years.

428
00:29:34,060 --> 00:29:35,660
It's got to be super hardened.

429
00:29:36,640 --> 00:29:50,500
Anyways, Opus found like 22 security vulnerabilities in a span of two weeks, which is like 30, 40% of all of the vulnerabilities that we found the entire previous year, which is amazing.

430
00:29:51,220 --> 00:29:58,120
And so they started like submitting patches, like submitting PRs to the code base.

431
00:29:58,270 --> 00:30:00,960
And we saw that and we were like, whoa, what is going on here?

432
00:30:01,140 --> 00:30:07,380
So we like reached out to that team and we've actually been working with them ever since we fixed all this stuff like right away.

433
00:30:08,060 --> 00:30:16,600
And then we've been working really closely with them because like it sort of like crossed this chasm where we were like, wow, actually, AI can do this stuff now.

434
00:30:16,780 --> 00:30:18,200
Like this is really powerful.

435
00:30:18,880 --> 00:30:25,660
Imagine how much stuff we could find and fix in the Firefox browser that are like serious security issues.

436
00:30:26,070 --> 00:30:29,820
And so we've gotten pretty close with that with that team over there, which is which is nice.

437
00:30:30,020 --> 00:30:33,220
They've let us like take some things for a spin and, you know, try, try new things.

438
00:30:33,280 --> 00:30:40,860
And obviously like they love working with us because we fix stuff really fast and we have some of the best security engineers on the planet.

439
00:30:41,220 --> 00:30:42,760
And so anyway, so that's been really fun.

440
00:30:42,980 --> 00:30:54,280
It's been like a fun thing to do with, uh, with that team, but impressive like use case for, uh, a browser that we were like, sure, we, you would never find anything in.

441
00:30:54,900 --> 00:31:11,060
Yeah. Given it doesn't, I haven't heard maybe they are planning something internally on like their own browser. But I think the fact they went for an extension, I think speaks to them maybe wanting something more open, which would be very nice. I feel like I don't get the vibe so far that they're going for just a locked in ecosystem for everything.

442
00:31:11,770 --> 00:31:12,420
I don't think so.

443
00:31:12,510 --> 00:31:22,120
So I don't know if you guys are open to anything, like if like there was a Firefox extension that they might be working on or anything to expect in the future.

444
00:31:22,460 --> 00:31:24,200
Yeah, we've talked about that with them.

445
00:31:24,300 --> 00:31:26,280
We've talked about just sort of building our own

446
00:31:26,440 --> 00:31:27,780
that looks like theirs.

447
00:31:28,280 --> 00:31:29,600
I don't think we're quite there yet.

448
00:31:30,280 --> 00:31:32,920
We're kind of focused on doing some other stuff with them,

449
00:31:33,060 --> 00:31:34,480
like a lot of the security stuff,

450
00:31:34,640 --> 00:31:36,000
which has been really helpful for us.

451
00:31:36,280 --> 00:31:37,860
So maybe eventually, I think.

452
00:31:37,920 --> 00:31:39,660
But yeah, I mean, we're very friendly with that team,

453
00:31:39,860 --> 00:31:40,960
which has been fun.

454
00:31:41,860 --> 00:31:42,840
And like I said,

455
00:31:43,080 --> 00:31:44,760
seem to be pretty values aligned with us,

456
00:31:44,880 --> 00:31:45,340
which is good.

457
00:31:45,820 --> 00:31:46,120
Very nice.

458
00:31:46,720 --> 00:31:49,300
Okay, now back to on topic.

459
00:31:49,680 --> 00:31:51,320
Back to our regularly scheduled programming.

460
00:31:52,140 --> 00:31:56,320
Yeah, so you mentioned the AI controls, and I wanted to kind of get into this next.

461
00:31:56,990 --> 00:31:57,080
Sure.

462
00:31:57,480 --> 00:32:00,680
So, okay, my understanding of events, this could be wrong.

463
00:32:01,430 --> 00:32:03,000
I keep up with too many articles back here.

464
00:32:03,010 --> 00:32:03,940
It's like hundreds a week.

465
00:32:04,760 --> 00:32:06,040
You guys released the AI features.

466
00:32:07,480 --> 00:32:11,800
Naturally, it was a polarized kind of assessment.

467
00:32:12,040 --> 00:32:12,860
People haven't even seen it yet.

468
00:32:13,010 --> 00:32:16,440
They haven't tried it yet, but it's still, you know, a lot of people are anti-AI, period.

469
00:32:17,060 --> 00:32:22,520
and then we saw the Firefox AI controls come out as kind of a separate announcement after the kind

470
00:32:22,560 --> 00:32:26,120
of a little bit of the backlash so do you mind maybe walking through what that looks like what

471
00:32:26,260 --> 00:32:31,460
the AI controls do what they don't do yeah so so first of all like we always knew we were going to

472
00:32:31,540 --> 00:32:37,900
build ways of turning stuff on and off that's like just a thing that we had as part of our plan

473
00:32:38,600 --> 00:32:43,639
what you saw was like us sort of testing some things with AI and then figuring out like okay

474
00:32:43,660 --> 00:32:45,400
what does this look like to turn it on and off?

475
00:32:45,620 --> 00:32:47,780
Now we have a meaningful number of features.

476
00:32:47,990 --> 00:32:49,640
So it makes sense to have that feature now.

477
00:32:49,650 --> 00:32:52,120
So it's sort of like part of a sequential roadmap.

478
00:32:53,160 --> 00:32:55,140
And since we can't release like everything all at once,

479
00:32:55,330 --> 00:32:58,120
we have to like build and ship and build and ship and build and ship.

480
00:32:58,500 --> 00:33:00,020
So that's like generally the backstory.

481
00:33:00,540 --> 00:33:03,920
Now how they work on the surface, they're really, really simple.

482
00:33:04,820 --> 00:33:07,100
But underneath the hood, they're actually quite complex.

483
00:33:07,230 --> 00:33:12,539
And there was like a lot of discussion and debate and like figuring out how's this going to work

484
00:33:12,560 --> 00:33:14,520
and like what states are we going to have

485
00:33:14,520 --> 00:33:15,180
and all this sort of stuff.

486
00:33:15,380 --> 00:33:18,400
So what it is on like the surface

487
00:33:18,760 --> 00:33:21,540
is it's a new area in Firefox settings.

488
00:33:21,930 --> 00:33:23,200
So you go into the browser settings

489
00:33:23,940 --> 00:33:25,760
and there's a big thing, this is AI controls.

490
00:33:26,980 --> 00:33:29,980
And you go into it and it lets you make decisions

491
00:33:30,420 --> 00:33:35,100
on whether AI is globally on or off

492
00:33:36,530 --> 00:33:39,120
or whether you want it off,

493
00:33:39,300 --> 00:33:41,539
but you want certain AI enhanced features

494
00:33:41,540 --> 00:33:47,540
to be available. So like an example might be translation. So a lot of people like translation.

495
00:33:48,200 --> 00:33:55,160
And we actually were originally just going to build the thing as a blanket on off toggle.

496
00:33:55,780 --> 00:33:59,380
It's either on or it's off. Simple. And then we talked to a bunch of users about it. And they

497
00:33:59,380 --> 00:34:07,480
were like, wait a minute. Like, I don't like AI, but I do like translations that are done with ML.

498
00:34:07,500 --> 00:34:11,040
And we were like, but that's kind of part of the AI umbrella.

499
00:34:11,500 --> 00:34:12,960
So like, what should we do?

500
00:34:13,060 --> 00:34:16,760
And they were like, well, could we turn most things off and some things back on?

501
00:34:16,810 --> 00:34:19,260
And we were like, sure, that seems reasonable.

502
00:34:19,700 --> 00:34:22,179
Like, so we're like, give the people what they want.

503
00:34:22,460 --> 00:34:29,280
So you can, if you're a no AI person, you can flip the switch, blanket, turns everything off.

504
00:34:29,490 --> 00:34:34,220
And like I mentioned earlier, it turns off like any mention of it.

505
00:34:34,460 --> 00:34:38,560
So like, we're not gonna show you a thing that's like,

506
00:34:38,860 --> 00:34:41,139
you should try smart window when that comes out.

507
00:34:41,340 --> 00:34:42,720
Like not happening, gone.

508
00:34:43,179 --> 00:34:45,020
Stricken from the record doesn't even exist

509
00:34:45,659 --> 00:34:47,040
like in the memory hole.

510
00:34:47,360 --> 00:34:49,120
And so you can do that if you block it all.

511
00:34:49,580 --> 00:34:52,100
But if you still wanna leverage some aspect

512
00:34:52,250 --> 00:34:53,360
like translation or you're like,

513
00:34:53,440 --> 00:34:54,320
I wanna turn off everything else,

514
00:34:54,370 --> 00:34:56,360
but I wanna test out the smart window thing.

515
00:34:56,820 --> 00:34:57,860
Like you can leave that on.

516
00:34:58,240 --> 00:35:00,140
When you block everything

517
00:35:00,350 --> 00:35:01,880
or you block an individual thing,

518
00:35:02,220 --> 00:35:03,600
we delete the models from your device.

519
00:35:04,120 --> 00:35:08,040
so like that was another step where we like should we do that and we're like look if people don't want

520
00:35:08,160 --> 00:35:15,820
ai they don't want it so get rid of it how much space are all of them together not not a ton i

521
00:35:15,900 --> 00:35:19,800
don't know the i don't know like how much they all are together but like not a ton these are these

522
00:35:19,800 --> 00:35:25,240
are small these are like small on-device models um and for performance reasons like we wouldn't

523
00:35:25,560 --> 00:35:30,040
ship the browser with like a huge amount of you know on-device stuff but it's also just more the

524
00:35:30,060 --> 00:35:34,200
principle of the thing. Like if I said, I don't want it, I don't want it. Like I don't want it

525
00:35:34,400 --> 00:35:39,100
anywhere on my system. So get it off. So we do that. Like they're really gone. Like I said,

526
00:35:39,100 --> 00:35:44,600
we don't suggest them. We delete the models like gonzo. And then as new features come online that

527
00:35:44,740 --> 00:35:48,960
take advantage of AI. And by the way, like not every, we're, we're not just shipping AI features

528
00:35:49,180 --> 00:35:53,920
every day. Like we're shipping a lot of things. Sometimes things are benefited by that technology,

529
00:35:54,520 --> 00:36:23,620
But as they come online and they have AI, like powering them or participating in or use the leveraging that technology, they'll go into the AI settings or the AI controls. And so that will grow over time. And we'll probably have to like, reorganize it and stuff like that. But it's still going to work the same. Turn it off, or leave it on, or turn off certain things or like whatever. Yeah, that's kind of it. That's that's pretty much how it works. But as you can imagine, under the hood, like,

530
00:36:24,120 --> 00:36:25,620
making sure all these things happen.

531
00:36:25,880 --> 00:36:26,840
Like, okay, I turned it off.

532
00:36:27,040 --> 00:36:27,820
The models are gone.

533
00:36:28,000 --> 00:36:29,400
Like what happens if somebody turns it off

534
00:36:29,420 --> 00:36:30,420
and turns it on really quick?

535
00:36:30,600 --> 00:36:31,800
Like, what do we do?

536
00:36:31,920 --> 00:36:32,940
We're like, well, the models are gone.

537
00:36:33,220 --> 00:36:33,340
Sorry.

538
00:36:33,640 --> 00:36:35,600
And by the way, just so folks know,

539
00:36:35,760 --> 00:36:38,220
like models don't download to the machine

540
00:36:39,040 --> 00:36:41,240
until somebody actually goes and engages

541
00:36:41,580 --> 00:36:43,240
with an AI feature in the first place.

542
00:36:43,380 --> 00:36:46,160
So it's not like we like load up a bunch of things

543
00:36:46,760 --> 00:36:48,340
in the browser package.

544
00:36:48,820 --> 00:36:51,600
Like they are downloaded at the moment

545
00:36:51,800 --> 00:36:53,980
when you're like, I would like to use this now.

546
00:36:54,480 --> 00:36:59,340
and so you know and then we're like okay this is somebody who wants the thing so here is the thing

547
00:36:59,760 --> 00:37:03,380
and it happens like behind the scenes and and stuff like that but then they're equally just

548
00:37:03,580 --> 00:37:08,540
deleted behind the scenes when you turn it off so yeah so that's how it works yeah very nice i have

549
00:37:08,580 --> 00:37:13,480
a few more technical questions about this before i i go into i guess what i've naturally grouped up

550
00:37:13,600 --> 00:37:19,300
here as we're talking as the case for ai skeptics yeah but before i get there which of these ai

551
00:37:19,500 --> 00:37:24,040
features or are all of them also going to be on mobile is this desktop only is it mobile we

552
00:37:24,060 --> 00:37:30,200
sort of selectively decide if things should be at parity based on the use case of the feature.

553
00:37:30,980 --> 00:37:38,940
So like, as an example, on iOS, we have this shake to summarize feature. And it's been my people are

554
00:37:39,060 --> 00:37:43,780
kind of into it. We obviously can't have a shake to summarize feature on the desktop, unless people

555
00:37:43,880 --> 00:37:48,060
are violently shaking their computer or something like that. So like, that's not a two way street.

556
00:37:48,720 --> 00:37:50,520
But controls is a two-way street.

557
00:37:50,640 --> 00:37:54,380
So like AI controls is coming to all the mobile experiences.

558
00:37:54,920 --> 00:37:57,420
A lot of people, when we rolled it out,

559
00:37:57,440 --> 00:37:58,880
were like, well, why isn't it there already?

560
00:37:59,080 --> 00:38:01,420
And we were like, well, we got to figure out,

561
00:38:01,660 --> 00:38:02,620
work out some of the kinks.

562
00:38:02,760 --> 00:38:04,460
And the easiest way to do that actually

563
00:38:04,720 --> 00:38:07,060
is to do it in one platform and then learn,

564
00:38:07,760 --> 00:38:09,800
especially the platform that has the most users,

565
00:38:10,000 --> 00:38:12,620
like learn and then roll it out to the other platforms

566
00:38:13,000 --> 00:38:16,180
because it makes the time it takes for us to go from,

567
00:38:16,340 --> 00:38:18,600
we'd like to release this to actually releasing it.

568
00:38:18,860 --> 00:38:22,420
much, much, much shorter. So that's an example where like something will be everywhere.

569
00:38:23,140 --> 00:38:27,640
And then there's other features and functionality that just like, maybe they make sense on mobile,

570
00:38:27,880 --> 00:38:31,860
maybe they don't. If we do some sort of like voice thing, maybe that makes sense on a mobile device

571
00:38:32,160 --> 00:38:36,820
where it's hard to type. Maybe it makes sense in the desktop browser, but like to be determined,

572
00:38:37,160 --> 00:38:43,180
I think. So we don't have like a hard and fast rule for must it be on both platforms. We kind of look

573
00:38:43,240 --> 00:38:48,039
at like, does it, does it make sense? One thing that I can tell you makes a ton of sense on mobile

574
00:38:49,160 --> 00:38:51,260
is tab grouping an organization

575
00:38:51,890 --> 00:38:54,060
because that is painful on a mobile device.

576
00:38:54,430 --> 00:38:55,760
Like no matter the platform,

577
00:38:56,480 --> 00:38:58,260
Mozilla aside, Firefox aside,

578
00:38:58,700 --> 00:39:00,860
like that's hard to manage a bunch of tabs.

579
00:39:01,360 --> 00:39:03,260
So if there's a way that AI can give you

580
00:39:03,500 --> 00:39:05,000
intelligent groupings and recommendations,

581
00:39:05,540 --> 00:39:08,140
that might be inherently really useful for people.

582
00:39:08,700 --> 00:39:10,300
And again, they don't have to do it.

583
00:39:10,300 --> 00:39:11,100
They can turn it off.

584
00:39:11,660 --> 00:39:13,340
But that's the thing that we're like thinking about.

585
00:39:13,440 --> 00:39:16,220
How do we bring that to a mobile experience?

586
00:39:16,620 --> 00:39:18,440
So yeah, so hopefully that answers your question.

587
00:39:19,160 --> 00:39:21,140
Yeah, and then, you know, I also was wondering,

588
00:39:22,000 --> 00:39:24,380
I feel like a lot of people are anti-I for privacy reasons.

589
00:39:25,060 --> 00:39:25,200
Yeah.

590
00:39:25,680 --> 00:39:27,620
So are all of these models offline?

591
00:39:27,980 --> 00:39:29,900
Do any of them have internet connected connectivity?

592
00:39:30,250 --> 00:39:31,520
Are they disclosed when they are?

593
00:39:31,820 --> 00:39:34,540
Most of the models download locally to your device.

594
00:39:35,300 --> 00:39:37,040
For something like Smart Window,

595
00:39:38,300 --> 00:39:42,720
where maybe you're choosing between a variety of different models,

596
00:39:43,360 --> 00:39:45,320
those have to be hosted somewhere.

597
00:39:45,560 --> 00:39:48,520
Some of them can be accessed locally, but some of them have to be hosted somewhere.

598
00:39:49,200 --> 00:39:55,260
And in that case, we're really clear about what you're sort of opting in to do.

599
00:39:55,480 --> 00:40:06,220
You're opting in for some things to be sent to a provisioned instance that we control that is secured for doing inference and things like that.

600
00:40:06,900 --> 00:40:10,220
But the main thing is we don't retain any of that stuff.

601
00:40:10,660 --> 00:40:18,400
So like if that if stuff does leave your device and we're transparent about like it has to leave the device here, we're not holding on to that information.

602
00:40:18,840 --> 00:40:21,440
One sort of violates our principles for stuff.

603
00:40:21,870 --> 00:40:25,480
And two, like, you know, we're not doesn't really benefit us in any way.

604
00:40:25,560 --> 00:40:26,720
Like, why would we want to hold that?

605
00:40:26,840 --> 00:40:27,820
That's not what users want.

606
00:40:28,140 --> 00:40:31,200
Like, why would we do something that our users objectively don't want?

607
00:40:31,360 --> 00:40:41,260
And so, you know, even if it's like a memory or a query that has to go somewhere to return an answer, like it's deleted after that session.

608
00:40:41,500 --> 00:40:42,660
It never stays.

609
00:40:42,820 --> 00:40:43,860
It never leaves that.

610
00:40:44,540 --> 00:40:47,220
And it's sent in very secure ways.

611
00:40:47,640 --> 00:40:51,940
So it's why we do things like OHDP and stuff like that in other parts of the browser, too.

612
00:40:52,020 --> 00:41:03,260
Like we often go to really great technical lengths to ensure that we're doing the right thing by our users and our manifesto, even if it's not the fastest engineering route.

613
00:41:03,450 --> 00:41:08,440
And so often a lot of people are like, why can't you just get this other feature out the door as quickly as possible?

614
00:41:08,450 --> 00:41:12,740
And we're like, well, it's complicated to do it the Mozilla way.

615
00:41:13,010 --> 00:41:14,300
And so, you know, it takes longer.

616
00:41:14,470 --> 00:41:18,620
But yeah, some things will have to leave device temporarily.

617
00:41:19,040 --> 00:41:24,140
But again, we're really transparent about that you're entering an experience where that could happen.

618
00:41:24,600 --> 00:41:26,800
And again, you don't have to go into that experience at all.

619
00:41:27,100 --> 00:41:29,640
So yeah, the remaining things are all on device.

620
00:41:30,420 --> 00:41:31,040
So yeah.

621
00:41:31,520 --> 00:41:31,660
Nice.

622
00:41:31,720 --> 00:41:35,320
And then my last technical question before we open up the skeptics section.

623
00:41:36,560 --> 00:41:48,060
Firefox forks, what part of this is, are all of these models and the AI features part of the same code base that potentially if someone was just forking the Firefox code and changing a few things,

624
00:41:48,620 --> 00:41:51,880
are they going to have all these AI features by default kind of baked in?

625
00:41:52,100 --> 00:41:55,540
Or how does this impact downstream forks of Firefox?

626
00:41:55,900 --> 00:41:58,140
Yeah, that's a really good question.

627
00:41:58,540 --> 00:42:00,960
I actually am not 100% sure of the answer to that.

628
00:42:01,190 --> 00:42:07,280
I would imagine things that have to hit the server access points and stuff that we control,

629
00:42:07,600 --> 00:42:12,040
probably that's not going to happen in forks.

630
00:42:12,400 --> 00:42:15,800
But I think the ability to pull down a model,

631
00:42:16,280 --> 00:42:19,060
like all that features and functionality should be there

632
00:42:19,340 --> 00:42:21,300
and running things locally should be there.

633
00:42:21,370 --> 00:42:22,960
But again, you know, people,

634
00:42:23,420 --> 00:42:25,040
one of the beautiful things about our browser

635
00:42:25,240 --> 00:42:28,080
is you can kind of customize it to work however you want.

636
00:42:28,180 --> 00:42:29,340
Like I, for example,

637
00:42:29,470 --> 00:42:32,820
have a locally running version of Firefox on my machine

638
00:42:33,120 --> 00:42:36,380
that I test out building new things in all the time.

639
00:42:36,820 --> 00:42:39,220
It's like a fun side project that I have.

640
00:42:39,480 --> 00:42:40,780
And it's pretty cool.

641
00:42:41,070 --> 00:42:43,720
And I mean, I don't use that as my daily driver browser.

642
00:42:43,960 --> 00:42:46,780
I use like the regular version of Firefox or nightly or something,

643
00:42:47,000 --> 00:42:49,020
but I have Adam's version of nightly.

644
00:42:49,260 --> 00:42:52,460
I was literally going to say, Adam, we need to see Adam's version.

645
00:42:52,860 --> 00:42:55,940
I had to, I had to label it because I was like,

646
00:42:56,060 --> 00:42:58,720
it looks so much like the other version of nightly.

647
00:42:58,800 --> 00:43:00,080
I'm like, how do I know which one I'm using?

648
00:43:00,600 --> 00:43:03,500
And so it literally in the dropdown, it says Adams Firefox nightly,

649
00:43:03,820 --> 00:43:08,700
but I've, I've like extended split view and like all kinds of things to just try.

650
00:43:08,720 --> 00:43:10,800
I'm like, what's possible. Could I do this? You know?

651
00:43:11,860 --> 00:43:13,260
And it's been really fun to do it.

652
00:43:13,620 --> 00:43:14,080
So, yeah.

653
00:43:14,900 --> 00:43:25,200
So, yeah, I wanted to just dedicate a section maybe to skeptics, mainly because I feel like maybe even a year ago, I would have been in this.

654
00:43:25,400 --> 00:43:26,600
This would have been my demographic.

655
00:43:26,880 --> 00:43:33,960
I've become a little more open to the idea of AI, especially now that I'm seeing the tools be a lot more useful for me personally.

656
00:43:34,120 --> 00:43:39,880
And also the use cases for the more private alternatives are starting to already get better.

657
00:43:41,580 --> 00:43:48,260
There's this relationship I see in privacy respecting technologies, which is big tech just does something.

658
00:43:48,400 --> 00:43:55,580
They don't care about the implications and the downstream effects, but they release this really cool thing that creates a lot of FOMO.

659
00:43:56,300 --> 00:43:57,360
And then it takes some time.

660
00:43:57,620 --> 00:44:04,200
And then finally, you know, people find a way to do it on device or they find a way to do it online with end-to-end encryption or zero knowledge, whatever.

661
00:44:04,580 --> 00:44:11,200
I feel like the AI gap has been not closed, but we're already starting to see what it looks like.

662
00:44:11,300 --> 00:44:17,100
So if someone is skeptical of AI or listening to this podcast, maybe they're in like, I'm always going to say no to AI camp.

663
00:44:18,360 --> 00:44:21,540
I had kind of two core questions to ask you for those people.

664
00:44:21,660 --> 00:44:28,860
The first one is, what is like an honest discussion on AI and what it can and can't do based on you slash Mozilla's philosophy?

665
00:44:29,300 --> 00:44:32,580
And do you see AI needing to be in their lives?

666
00:44:32,880 --> 00:44:39,480
Or is this something that, I know it's kind of a difficult question to answer, but I would start there and then I have a second question.

667
00:44:39,600 --> 00:44:46,140
Yeah. So like honest discussion on AI, what it can and can't do. What you just said, which was,

668
00:44:46,290 --> 00:44:50,220
you know, a year ago, you were a skeptic and now you're starting to warm up because you see how

669
00:44:50,520 --> 00:44:56,920
it's becoming more useful. I tend to agree with that. Like my anthropic example of like a year

670
00:44:57,010 --> 00:45:02,280
ago, we tried a thing or six, eight, nine months ago, we tried a thing and it failed miserably.

671
00:45:02,900 --> 00:45:06,820
And then December, January, guess what?

672
00:45:07,210 --> 00:45:08,740
It works really well now.

673
00:45:09,300 --> 00:45:15,200
And that pace is fast, how good some of these things are getting.

674
00:45:15,680 --> 00:45:21,220
So the first thing I would say to skeptics is you don't have to love AI.

675
00:45:21,640 --> 00:45:23,960
It's actually good to be skeptical of it.

676
00:45:23,960 --> 00:45:25,180
I teach this to my kids.

677
00:45:26,170 --> 00:45:27,540
Be skeptical of the output.

678
00:45:28,720 --> 00:45:31,480
Independent thought is not gone in society.

679
00:45:32,080 --> 00:45:34,320
and it shouldn't be, at least like in my family.

680
00:45:34,660 --> 00:45:35,940
I'm like, that's really important.

681
00:45:36,020 --> 00:45:37,680
Our ability to think critically about stuff

682
00:45:37,820 --> 00:45:40,640
and to interrogate things that we get back from AI

683
00:45:41,020 --> 00:45:42,340
is really, really important.

684
00:45:43,180 --> 00:45:45,020
Like by the time that this show goes out,

685
00:45:45,740 --> 00:45:47,300
what AI can and can't do

686
00:45:47,440 --> 00:45:48,960
will probably have changed again, right?

687
00:45:49,200 --> 00:45:51,280
Like I don't even know what the release schedule is,

688
00:45:51,460 --> 00:45:55,100
but the pace of which things are improving is wild.

689
00:45:55,660 --> 00:46:00,240
So I would say like things that AI can do today pretty well.

690
00:46:00,640 --> 00:46:05,120
Obviously, those coding use cases that I gave examples of are like, it's pretty impressive.

691
00:46:05,620 --> 00:46:11,100
Still requires human review, but it gets people a lot farther along a lot faster.

692
00:46:11,520 --> 00:46:14,940
It's allowing a lot of companies to build products faster for people.

693
00:46:15,400 --> 00:46:18,180
That's genuinely good as long as they're building the right stuff.

694
00:46:18,760 --> 00:46:28,700
I think aside from that specific case, AI can take information and make it easier and more accessible to people and help them learn and grow over time.

695
00:46:29,780 --> 00:46:39,480
It is generally a good teaching tool to ask it questions, get feedback, ask that feedback questions, like continue to go and turn those conversations.

696
00:46:40,120 --> 00:46:41,260
It's interesting.

697
00:46:42,040 --> 00:46:53,940
Democratizing like learning, you know, like my sometimes the AI stuff that I generate from my daughter's math test is just as helpful as the tutor that we pay for.

698
00:46:53,940 --> 00:46:57,180
And in fact, those two things working together are actually way more beneficial.

699
00:46:57,620 --> 00:47:11,840
So like, I think eventually, it's going to help people get a lot of things done that are kind of drudgery tasks, repetitive, formulaic, it's going to save people time, and they're going to not have to do the stuff that they don't like to do anymore.

700
00:47:12,120 --> 00:47:18,160
Today, you have to have like a corpus of really high quality information and context to do that.

701
00:47:18,430 --> 00:47:27,040
And it still fails when there isn't enough context or high quality information or it doesn't know something.

702
00:47:27,920 --> 00:47:31,460
It's not great at recognizing, oh, I don't actually know that.

703
00:47:31,650 --> 00:47:34,580
So I should tell Adam that or Henry that.

704
00:47:35,190 --> 00:47:36,380
Instead, it wants to please.

705
00:47:37,110 --> 00:47:38,780
So it's going to try to give you an answer.

706
00:47:39,130 --> 00:47:40,020
And that answer is wrong.

707
00:47:40,460 --> 00:47:45,220
you know, and like completely manufactured. That's happening less and less because I think

708
00:47:45,920 --> 00:47:50,660
sometimes some people are getting better at giving it more context. It's also gobbling up more

709
00:47:50,840 --> 00:47:55,900
information and stuff like that. But like broadly speaking, it still makes those kinds of mistakes.

710
00:47:56,380 --> 00:48:03,000
And so like we've spent a lot of time with our work on things like Smart Window that like if it

711
00:48:03,400 --> 00:48:08,319
objectively can't answer a question, it's going to let you know that it needs to search for the

712
00:48:08,340 --> 00:48:14,720
information like elsewhere, or it's going to be like, I don't have the answer to that question.

713
00:48:15,380 --> 00:48:19,720
Sorry. I think, let me tell you one more thing that it can't do. Because I told you some of the

714
00:48:19,730 --> 00:48:24,000
things I think it can do really well, like coding, information synthesis, and stuff like that. And

715
00:48:24,000 --> 00:48:29,060
you mentioned big tech kind of just like flinging things out of the ether and letting everybody else

716
00:48:29,280 --> 00:48:35,299
catch up. Let's talk about agentic workflows. These are things that AI cannot do super well

717
00:48:35,320 --> 00:48:42,240
right now, like complex agentic workflows. And by agentic, I mean it acting on your behalf to make a

718
00:48:42,360 --> 00:48:46,960
bunch of decisions, click on a bunch of things, like do a bunch of stuff. Like we at Firefox

719
00:48:48,120 --> 00:48:55,480
know that if we're going to do anything in that area, it has to be safe and secure. And it can't

720
00:48:55,620 --> 00:49:01,279
just go off into the ether and like do a bunch of things on your behalf that you didn't intend it to

721
00:49:01,300 --> 00:49:08,480
do. And so we are treading really lightly in that space. Like Smart Window at its launch is not

722
00:49:08,500 --> 00:49:15,900
going to have any of that stuff in it. Like if we do that at some point, it will be very specific

723
00:49:16,120 --> 00:49:22,600
use cases that are like heavily tested and pressure tested because like the more complicated

724
00:49:22,640 --> 00:49:27,359
a workflow is, the more opportunities an agent has to get it wrong. It's like that game of telephone

725
00:49:27,360 --> 00:49:28,740
that we all used to play in school.

726
00:49:29,370 --> 00:49:30,640
Like I like whisper in your ear

727
00:49:30,820 --> 00:49:32,140
and you whisper in somebody's else ear

728
00:49:32,320 --> 00:49:33,440
and 10 steps down the road,

729
00:49:33,600 --> 00:49:35,360
it's not even the same like sentence

730
00:49:35,570 --> 00:49:36,420
that we started with.

731
00:49:36,920 --> 00:49:39,060
Like that's agentic workflows right now.

732
00:49:39,580 --> 00:49:40,580
And so we don't like,

733
00:49:41,419 --> 00:49:43,960
it's irresponsible for us to like purport

734
00:49:44,060 --> 00:49:45,300
that that works well for users.

735
00:49:45,800 --> 00:49:46,600
And so like, you're not,

736
00:49:46,950 --> 00:49:47,940
like that's a thing that,

737
00:49:48,160 --> 00:49:49,740
you know, some companies are saying like,

738
00:49:49,820 --> 00:49:50,360
oh, it works great.

739
00:49:50,560 --> 00:49:51,300
It's awesome.

740
00:49:52,060 --> 00:49:52,740
Yeah, use it.

741
00:49:52,880 --> 00:49:53,500
Use all the things.

742
00:49:53,610 --> 00:49:54,080
And we're like,

743
00:49:55,220 --> 00:49:56,520
we're not so sure.

744
00:49:57,200 --> 00:50:00,720
So like, that's not going to find its way into the browser anytime soon.

745
00:50:00,970 --> 00:50:04,440
So like stuff like that, like I think it again, I'll come all the way back to the beginning.

746
00:50:04,580 --> 00:50:06,860
Like, it's good to be skeptical of this stuff.

747
00:50:06,990 --> 00:50:08,140
I actually encourage it.

748
00:50:08,430 --> 00:50:10,380
I love talking to AI skeptics.

749
00:50:10,900 --> 00:50:13,140
And I like talking to AI true believers, too.

750
00:50:13,240 --> 00:50:18,340
I had a friend from OpenAI at my house one time, and I was just grilling him.

751
00:50:18,690 --> 00:50:19,980
I was like, are we in a bubble?

752
00:50:20,420 --> 00:50:21,440
What's going on here?

753
00:50:21,660 --> 00:50:22,540
How is this useful?

754
00:50:22,860 --> 00:50:24,400
Tell me how, like, all these things.

755
00:50:24,680 --> 00:50:26,080
And like, it was a good conversation.

756
00:50:26,320 --> 00:50:27,820
And I think you should do that sort of stuff.

757
00:50:28,020 --> 00:50:30,760
Like it only makes people better.

758
00:50:30,980 --> 00:50:34,360
It makes this stuff better and gets the outcome that we want.

759
00:50:34,760 --> 00:50:36,880
So anyway, that's a long answer to your question,

760
00:50:37,000 --> 00:50:41,080
but that's how I think about it for skeptics and what it can and can't do.

761
00:50:41,520 --> 00:50:43,380
Yeah, and I want to be mindful of time.

762
00:50:44,140 --> 00:50:46,260
I only have a couple more, but as you're talking,

763
00:50:46,720 --> 00:50:49,340
and you don't have to go super deep into it either,

764
00:50:49,600 --> 00:50:51,740
but I was kind of thinking like,

765
00:50:51,900 --> 00:50:55,640
what are the friction points I've seen from different people

766
00:50:55,640 --> 00:50:56,980
who don't want to use AI.

767
00:50:57,150 --> 00:50:57,980
And I think one of them is

768
00:50:58,010 --> 00:50:59,120
they just don't see the use case,

769
00:50:59,440 --> 00:51:00,640
which is kind of what you touched on.

770
00:51:00,780 --> 00:51:03,380
But another thing that came up is ethics.

771
00:51:04,600 --> 00:51:05,820
I've seen a lot of people concerned

772
00:51:06,000 --> 00:51:07,460
about how data is trained,

773
00:51:07,920 --> 00:51:11,400
how most of the companies involved

774
00:51:12,260 --> 00:51:13,760
in this space are all,

775
00:51:14,780 --> 00:51:16,300
in some way, for the most part,

776
00:51:16,700 --> 00:51:17,560
unethical companies

777
00:51:17,730 --> 00:51:19,920
based on these people's views,

778
00:51:20,500 --> 00:51:21,580
including mine sometimes.

779
00:51:23,280 --> 00:51:24,280
The privacy implications.

780
00:51:24,960 --> 00:51:29,720
the IP implications, the environmental implications potentially as well.

781
00:51:29,720 --> 00:51:33,660
I know that seems to be changing every day if AI actually uses a ton of water or not,

782
00:51:33,900 --> 00:51:34,700
depending on who you ask.

783
00:51:35,920 --> 00:51:39,780
So, you know, data centers, people are very against data centers being built.

784
00:51:40,280 --> 00:51:43,260
People's energy prices are going up because of AI.

785
00:51:43,800 --> 00:51:47,020
Like, these are the things that also, I think, make people go,

786
00:51:47,300 --> 00:51:48,520
I don't want to use this technology.

787
00:51:48,550 --> 00:51:51,100
I don't want to support this industry in any capacity.

788
00:51:52,640 --> 00:51:57,980
Is this maybe more so where we go back to what you said about showing up and showing an alternative way of doing things?

789
00:51:58,090 --> 00:51:59,620
Is that how you would see this personally?

790
00:52:00,340 --> 00:52:05,840
Yeah, I think Mozilla will always do as much as we can across a lot of those dimensions.

791
00:52:06,170 --> 00:52:20,860
So if you take ethical training, we have the Common Voice Project, which is an open source way that people can actually donate their voice to do language transcription, like voice transcription.

792
00:52:21,180 --> 00:52:26,520
And you come into it and you know what you're doing and you're like helping translate things

793
00:52:26,700 --> 00:52:27,920
and you're speaking things back.

794
00:52:27,960 --> 00:52:30,880
And it's like, I'm donating my voice and my time to this effort.

795
00:52:31,300 --> 00:52:37,460
And it's like an open repository of training data that everyone is like, we're all in this

796
00:52:37,660 --> 00:52:37,780
together.

797
00:52:37,900 --> 00:52:38,640
You can use this.

798
00:52:38,980 --> 00:52:43,560
And yeah, we're like pretty, when I think about like policy and regulation and stuff

799
00:52:43,760 --> 00:52:47,600
like that, like, again, I mentioned that it's pretty early still on a lot of these areas,

800
00:52:48,300 --> 00:53:02,620
But Mozilla has been like elbow to elbow with the big, the big tech companies when it comes to other types of regulation, like anti-competitive stuff and choice screens in Europe and like digital, the Digital Markets Act and things like that.

801
00:53:02,630 --> 00:53:04,300
Like AI is not going to be any different.

802
00:53:04,510 --> 00:53:06,040
We're in all those conversations.

803
00:53:06,330 --> 00:53:11,160
We like go to the symposium in India where they're like having conversations about this.

804
00:53:11,160 --> 00:53:15,780
Like we're in the mix and we show up pretty big when we show up at these, these things.

805
00:53:16,060 --> 00:53:20,060
And so like we're trying to the best of our abilities.

806
00:53:20,490 --> 00:53:25,340
I think there are some things that like we have to be realistic about, like can Mozilla

807
00:53:25,640 --> 00:53:27,520
itself solve that problem?

808
00:53:28,220 --> 00:53:32,340
And I think there are some things that are like systemic ecosystem challenges that we

809
00:53:32,440 --> 00:53:34,460
can certainly advocate and be a voice.

810
00:53:34,950 --> 00:53:38,940
But like it's going to take a lot more than just us to push that forward.

811
00:53:39,410 --> 00:53:44,680
And so we try to like change the standards so that then everyone else has to do the same

812
00:53:44,840 --> 00:53:45,040
thing.

813
00:53:45,400 --> 00:53:58,500
And that's like the best thing that we can do is we can change the system so that you can't work around the system, you know, and that there are rules and there is like an expectation of what you do here and what energy usage and things like that.

814
00:53:58,630 --> 00:54:01,860
And so I don't think we can fix everything, but we're sure as hell trying.

815
00:54:02,990 --> 00:54:03,400
So, yeah.

816
00:54:03,920 --> 00:54:11,060
And then going back to my actual second question, which was, you know, the first one is what if people just don't see the use case, et cetera.

817
00:54:11,180 --> 00:54:15,120
I think the other concern is there are probably things that Mozilla does.

818
00:54:15,300 --> 00:54:18,880
Everyone uses a Mozilla product, whether it's Firefox,

819
00:54:19,220 --> 00:54:22,180
whether it's something else you guys have, for a particular reason.

820
00:54:22,920 --> 00:54:26,760
And I think potentially another skepticism someone might have is,

821
00:54:27,200 --> 00:54:31,200
oh, is this Mozilla shifting away from their priorities?

822
00:54:32,360 --> 00:54:37,140
I only want, and I respect Mozilla for their third-party web engine.

823
00:54:37,700 --> 00:54:39,060
I don't care about this AI stuff.

824
00:54:39,480 --> 00:54:41,240
and now they're going to be focusing on AI stuff.

825
00:54:41,400 --> 00:54:42,840
So should I start looking for another browser?

826
00:54:43,040 --> 00:54:46,560
Like I can see this be a pretty realistic chain of thoughts.

827
00:54:47,100 --> 00:54:47,920
So I was just curious,

828
00:54:48,160 --> 00:54:50,200
like what are your guys' priorities going forward

829
00:54:50,500 --> 00:54:51,580
and is that going to stay the same?

830
00:54:51,940 --> 00:54:53,300
Yeah, I can tell you that AI,

831
00:54:53,540 --> 00:54:55,180
if you look at like the number of people

832
00:54:55,420 --> 00:54:58,880
working on this at Mozilla is like this big

833
00:54:59,300 --> 00:55:01,320
relative to all the other non-AI things

834
00:55:01,480 --> 00:55:03,200
that the entire organization is working on.

835
00:55:03,340 --> 00:55:07,640
It's like maybe 5% of the company in some capacity

836
00:55:07,660 --> 00:55:12,700
spend some part of their day like working on something like this like like our ai product

837
00:55:12,900 --> 00:55:18,800
management team is like four people pretty small in like a thousand person company and so i think

838
00:55:18,900 --> 00:55:22,680
you know i think people are always going to push for the thing that they think is the most important

839
00:55:22,750 --> 00:55:28,260
thing to work on but i would say the teams that are working on the gecko rendering engine

840
00:55:29,530 --> 00:55:35,640
web compatibility ensuring that the firefox works on the most websites on the planet that it's

841
00:55:35,660 --> 00:55:40,680
perform it, like that actually still makes up the lion's share of the people at the company.

842
00:55:41,520 --> 00:55:46,500
And I don't see that changing anytime soon because like we're nothing if the browser doesn't work.

843
00:55:47,100 --> 00:55:52,360
And the Gecko rendering engine, like we're the we're like the last, the last of the great rendering

844
00:55:52,820 --> 00:55:59,080
engines, you know, it's like Chromium WebKit and us. And like, we can't walk away from that. Like,

845
00:55:59,160 --> 00:56:04,260
that's what makes us independent. That's what makes us like have a voice and have a seat at the table.

846
00:56:04,820 --> 00:56:08,680
And so like that stuff's going to remain super, super important for us forever.

847
00:56:08,980 --> 00:56:10,440
So that's what I would say about that.

848
00:56:10,580 --> 00:56:14,900
Like people will always have their sort of like lead horse that they're really excited about.

849
00:56:15,020 --> 00:56:17,900
And they wish that every one of the company was working on that one thing.

850
00:56:18,160 --> 00:56:19,540
Some people probably feel that way about Pocket.

851
00:56:19,760 --> 00:56:22,240
Like, could you just work on Pocket all of you all the time?

852
00:56:22,500 --> 00:56:24,520
And the reality is it doesn't work out that way.

853
00:56:25,020 --> 00:56:29,580
But I can assure people that like the majority of our energy and effort is being spent on

854
00:56:29,600 --> 00:56:36,680
building like a world class browser that works for as many people as possible. And we think AI

855
00:56:37,120 --> 00:56:41,640
is a component of that, but it's not, it's not the only meal. It's not the only dish on the menu.

856
00:56:42,000 --> 00:56:44,640
So to, so to speak. And that doesn't feel like it's changing here.

857
00:56:44,960 --> 00:56:49,920
Yeah. And then I had a Bobby Holly on from your team as well, not too long ago to talk about kind

858
00:56:50,040 --> 00:56:54,560
of the, the web kit or not the web kit, the gecko side of things and how it compares to web kit and

859
00:56:54,700 --> 00:56:59,560
chromium and stuff like this. So Bobby is one of the guys I work with on a daily basis. He's

860
00:56:59,580 --> 00:56:59,900
It's a good interview.

861
00:57:00,460 --> 00:57:01,740
So if you guys haven't listened to that,

862
00:57:01,920 --> 00:57:02,860
that was a good one too.

863
00:57:03,280 --> 00:57:03,520
Cool.

864
00:57:04,780 --> 00:57:05,620
My last question.

865
00:57:06,680 --> 00:57:08,280
It's kind of on the note of the skeptics,

866
00:57:08,480 --> 00:57:11,620
but I think it could just be a good finishing point for us here.

867
00:57:11,820 --> 00:57:15,520
How would you recommend someone who's confused about all of this get started?

868
00:57:16,080 --> 00:57:17,940
They're like, okay, I hear this interview.

869
00:57:18,620 --> 00:57:19,360
I've had it, you know,

870
00:57:19,480 --> 00:57:21,340
maybe I've had my AI controls off for everything

871
00:57:21,460 --> 00:57:22,260
because I don't want to touch it.

872
00:57:22,760 --> 00:57:23,780
Is there a specific feature

873
00:57:24,070 --> 00:57:26,500
or maybe is it something away from Firefox entirely

874
00:57:26,520 --> 00:57:29,660
to kind of like dip their toes into the AI world

875
00:57:29,820 --> 00:57:31,400
if they want to at least see if it's for them.

876
00:57:31,960 --> 00:57:35,680
Yeah, I mean, I would say there's a few things.

877
00:57:35,960 --> 00:57:38,180
Like one is like watching videos like this,

878
00:57:38,720 --> 00:57:41,120
hearing somebody like you talk about like,

879
00:57:41,960 --> 00:57:46,400
I'm skeptical, like convince me, why should I use this?

880
00:57:46,560 --> 00:57:48,020
Like just getting over that hump

881
00:57:48,040 --> 00:57:49,220
and sort of educating yourself

882
00:57:49,500 --> 00:57:51,600
is like a good starting point, I think for people.

883
00:57:51,960 --> 00:57:54,800
In terms of like specific features in the browser,

884
00:57:55,360 --> 00:57:57,820
You know, trying out Shake to summarize is kind of fun.

885
00:57:58,060 --> 00:57:59,300
It's going to be coming to Android.

886
00:57:59,620 --> 00:58:00,340
So that'll be a thing.

887
00:58:01,100 --> 00:58:04,580
So then it'll be available on as many platforms as possible.

888
00:58:05,480 --> 00:58:09,600
I mean, I think what we see most often for people who are turning it off is they turn

889
00:58:09,600 --> 00:58:13,580
it off and they leave translation on or they leave alt text generation on.

890
00:58:13,880 --> 00:58:16,940
Then it's like, well, are you trying AI at that point?

891
00:58:16,980 --> 00:58:18,860
Or is it kind of just working behind the scenes?

892
00:58:19,080 --> 00:58:22,980
It's probably more working behind the scenes, but at least you're getting some of the benefit

893
00:58:23,100 --> 00:58:23,480
of it.

894
00:58:23,680 --> 00:58:38,940
When it comes to like actually trying it out, I would say you might wait until Smart Window is live and know that Firefox and Mozilla are handling this in a way that is as aligned with your interests as any other, like better than any other company on the planet.

895
00:58:39,740 --> 00:58:51,700
And, you know, and then try like if you're more technical, try like a bring your own model kind of situation where, you know, maybe the models on your device or whatever, or try one of the open source models or, or something like that.

896
00:58:52,400 --> 00:58:55,080
And like, you know, just ask it some questions.

897
00:58:55,350 --> 00:58:57,400
I find that like a lot of times

898
00:58:58,020 --> 00:59:00,560
when I don't know what to do with AI

899
00:59:00,690 --> 00:59:02,220
or when I come into like a blank canvas,

900
00:59:02,370 --> 00:59:04,840
I'm kind of like, well, what can I do here?

901
00:59:05,080 --> 00:59:06,420
And it's like, oh, I have a bunch of ideas

902
00:59:06,580 --> 00:59:07,380
for things you can do.

903
00:59:07,530 --> 00:59:09,080
And it's like, oh, let me try one of those.

904
00:59:09,450 --> 00:59:11,160
And I just sort of do a little trial and error.

905
00:59:11,340 --> 00:59:12,080
And I kind of learned,

906
00:59:12,200 --> 00:59:14,020
that's how I've learned how to use Cloud Code.

907
00:59:14,280 --> 00:59:15,600
That's how I've learned how to use

908
00:59:15,930 --> 00:59:18,100
all the AI that I use in my life.

909
00:59:18,270 --> 00:59:21,059
But like you think of AI as like a plucky intern

910
00:59:21,540 --> 00:59:23,660
that's sort of like always willing to help, you know?

911
00:59:24,120 --> 00:59:26,520
And like, that's the good construct for like,

912
00:59:26,540 --> 00:59:29,320
well, what kind of a task could I ask this thing to do for me?

913
00:59:29,440 --> 00:59:30,400
Or how could it help me?

914
00:59:30,540 --> 00:59:32,760
Like intern level task, you know?

915
00:59:32,760 --> 00:59:33,200
That's a good way to put it.

916
00:59:33,460 --> 00:59:35,580
Expect it to get it wrong a little bit,

917
00:59:35,960 --> 00:59:39,300
but like generally it's gonna kind of help you

918
00:59:39,440 --> 00:59:40,720
in your day to day, you know?

919
00:59:40,920 --> 00:59:42,540
And so like, you know, drafting things,

920
00:59:42,840 --> 00:59:44,680
summarizing things, like taking dense,

921
00:59:45,520 --> 00:59:47,700
convoluted information and making sense out of it.

922
00:59:47,860 --> 00:59:50,100
Like, especially where people miss nuances

923
00:59:50,100 --> 00:59:51,940
across like really complicated,

924
00:59:52,750 --> 00:59:53,920
like chunks of information.

925
00:59:54,480 --> 00:59:56,060
It's actually really good at that stuff.

926
00:59:56,240 --> 00:59:59,160
Like, like, because that's what machines are good at.

927
00:59:59,460 --> 01:00:01,520
Like they're, they're better, they're better than us.

928
01:00:01,840 --> 01:00:03,420
And then I would say the other thing is like,

929
01:00:03,740 --> 01:00:04,840
don't lose the skepticism.

930
01:00:05,480 --> 01:00:08,480
Like AI is not a substitute for your own judgment

931
01:00:09,140 --> 01:00:10,800
or facts or accountability.

932
01:00:11,520 --> 01:00:14,680
Like even at Mozilla, we are really clear

933
01:00:14,910 --> 01:00:16,940
even about employees usage of AI.

934
01:00:17,620 --> 01:00:20,400
Like it's not a substitute for your judgment.

935
01:00:20,680 --> 01:00:26,960
Like we expect you are in control of the things that you take out of AI.

936
01:00:27,920 --> 01:00:31,560
And obviously we have really strict policies on what can be used and stuff there.

937
01:00:31,700 --> 01:00:34,260
But like it's not a replacement for human judgment.

938
01:00:34,680 --> 01:00:38,820
And it's not a replacement for like accountability if you're using it.

939
01:00:38,820 --> 01:00:39,760
So be skeptical.

940
01:00:40,140 --> 01:00:42,460
Like if you are skeptical, stay skeptical.

941
01:00:42,960 --> 01:00:44,340
Like approach it with that lens

942
01:00:44,840 --> 01:00:46,300
because I'm not gonna tell people

943
01:00:46,310 --> 01:00:47,420
to be different than they are.

944
01:00:47,800 --> 01:00:49,600
It's how you're wired, you know?

945
01:00:49,760 --> 01:00:51,020
And so like, I think that's important

946
01:00:51,320 --> 01:00:53,900
for people to approach it with the lens that they bring

947
01:00:54,620 --> 01:00:56,580
or if they just don't want to, don't approach it at all.

948
01:00:56,730 --> 01:00:58,360
But I think you'll be missing out

949
01:00:58,620 --> 01:01:01,380
if you don't like dip your toe in the water, as you said.

950
01:01:01,960 --> 01:01:02,040
Yeah.

951
01:01:02,160 --> 01:01:02,360
Got it.

952
01:01:03,170 --> 01:01:05,580
Do you have any place to send people

953
01:01:05,860 --> 01:01:07,420
to follow what you're working on,

954
01:01:07,770 --> 01:01:10,420
either on Mozilla's end or your personal end

955
01:01:10,730 --> 01:01:11,340
for people in the connect?

956
01:01:11,380 --> 01:01:24,820
My personal end has very little to do with Mozilla work, but I would say definitely we're trying to release as much as we can on the Mozilla blog and especially forward-looking roadmap type stuff.

957
01:01:25,420 --> 01:01:28,900
We are often teasing the things that are coming in future features.

958
01:01:29,760 --> 01:01:33,360
Mozilla Connect is also a great way to get involved.

959
01:01:33,570 --> 01:01:39,520
We have a community of people that are just clamoring to get access to stuff before it's available for the public.

960
01:01:39,700 --> 01:01:43,480
Like, we love that because it's often filled with some skeptics.

961
01:01:43,680 --> 01:01:47,380
And like, we want to hear all of the things that they don't like about it and like what

962
01:01:47,480 --> 01:01:48,200
the problems are.

963
01:01:48,360 --> 01:01:50,700
Like, that's actually the way that we build better products.

964
01:01:50,780 --> 01:01:55,320
So like that, I know you don't hang out on Reddit for obvious reasons, but like Reddit,

965
01:01:55,680 --> 01:01:56,960
we pay attention to that stuff.

966
01:01:57,020 --> 01:01:57,860
Like we really do.

967
01:01:58,320 --> 01:02:02,120
So any of the community touch points, socials, things like that, like those are the ways to

968
01:02:02,240 --> 01:02:03,520
kind of send us feedback.

969
01:02:03,740 --> 01:02:07,140
And we really do listen and pay attention to all of that stuff.

970
01:02:07,680 --> 01:02:09,160
We can't always act on it right away.

971
01:02:09,720 --> 01:02:15,740
But like, it's a big part of the inputs into how we think about building products and like

972
01:02:16,140 --> 01:02:22,300
understanding what users of the browser want and don't want. So keep doing all those things

973
01:02:23,140 --> 01:02:27,580
would be my advice to people. Very cool. Well, thank you for your time, Adam. I know we went

974
01:02:27,590 --> 01:02:33,900
a little over here, but I really appreciated the talk. And I think we did a good job diving into all

975
01:02:33,920 --> 01:02:39,660
the nuances of these AI features, hopefully. Yeah, I sure hope we did. And thanks, Henry,

976
01:02:39,700 --> 01:02:43,200
for having me on. This is really fun. You're a great interviewer. So I love talking about this

977
01:02:43,340 --> 01:02:48,120
stuff. So I'll come back anytime. Cool. Thanks, Adam. All right. And with all of that said,

978
01:02:48,360 --> 01:02:52,560
I just want to thank Adam for coming on the podcast, taking the time to be here and helping

979
01:02:52,900 --> 01:02:59,140
explain a little bit of the philosophy behind AI, the pros and cons and Mozilla's approach and where

980
01:02:59,140 --> 01:03:03,460
they think they are in the equation. If you have any more questions, definitely let me know.

981
01:03:03,560 --> 01:03:07,940
I probably will have them on for another interview down the road, especially regarding their new VPN

982
01:03:08,280 --> 01:03:12,560
product. So if you have any questions about their free VPN that is now inside of the Firefox browser,

983
01:03:12,760 --> 01:03:16,080
definitely let me know and I'm going to try to follow up with them and get that interview going.

984
01:03:16,420 --> 01:03:19,680
Last thing, if you like these podcasts and you want to support them and keep them growing,

985
01:03:20,080 --> 01:03:23,960
you can become a Techlorian in the description. This is a really easy way to get access to

986
01:03:24,200 --> 01:03:28,500
exclusive perks in our community like our private signal group. You also get access to some credits

987
01:03:28,800 --> 01:03:32,820
across both our website and our videos if you're into that kind of thing. And of course, you get

988
01:03:32,840 --> 01:03:34,000
to keep this podcast growing.

989
01:03:34,260 --> 01:03:35,060
That is all I have to say.

990
01:03:35,190 --> 01:03:36,380
I hope you learned something new today

991
01:03:36,510 --> 01:03:38,560
to keep yourself and the people around you safer.

992
01:03:38,730 --> 01:03:40,760
And I'll see you next time on Techlore Talks.

993
01:04:09,940 --> 01:04:09,960
.