Subscribe
Share
Share
Embed
Trying to keep up as AI accelerates toward escape velocity? Let the Mission Control crew help. Tune in daily for a quick take on the last 24 hours in AI news.
Adam:
It's Wednesday, June 7th. This is Accelerate Daily. Today we've got a look at the possibility of Apple's Vision Pro reading your mind. Sam Altman argues that open AI should remain public so it can get strange. Google is dropping free knowledge about how to use AI. And we've got a prompt engineering workshop looking at the use of personas to juice your LLM outputs. Get your hand away from that abort button. It's time for Accelerate Daily. music but I don't need a soundboard Okay. Welcome back everybody. I'm Adam.
Mackenzie:
I'm Mackenzie, good morning.
Adam:
And we're back with three headlines and one how to keep you caught up on what's happening in AI acceleration today. Uh, Mac, what are we looking at for today's AI image of the day?
Mackenzie:
Uh, that is a Valkyrie shopping. And
Adam:
Hahaha
Mackenzie:
what do you call these? A CVS in America? This looks like a shoppers
Adam:
Hobby
Mackenzie:
drug
Adam:
Lobby.
Mackenzie:
mart hobby
Adam:
It's
Mackenzie:
lobby.
Adam:
a Hobby Lobby. Yeah, let me cut over to the to the links. Take a look at the whole series. This is again from r slash mid journey. We need to talk about what's going on at Hobby Lobby. Won't someone please think of the children? Um, this is a series of images generated where, uh, with, with generally demonic things showing up in the aisles of
Mackenzie:
Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe
Adam:
Hobby Lobby, which for anyone lacking the context is a U S company that went to the mat to protect their right to not. Cover contraception for their. They're very Christian,
Mackenzie:
Hehehehe
Adam:
to the point that they went to the Supreme Court to get a religious exemption from paying for birth control pills under their insurance plan. Anyway, it's a better example though of the extent to which these are convincing images that could raise a furor about what's Hobby Lobby doing? Like
Mackenzie:
Yeah, what are
Adam:
completely...
Mackenzie:
all these Satans doing in
Adam:
Completely,
Mackenzie:
there?
Adam:
what are all these Satans doing in your otherwise publicly pious stores?
Mackenzie:
I'm sorry.
Adam:
Less silly than some of our other options, but I like to flag the stuff that's, you know, can get misinformation.
Mackenzie:
Yeah, this is what states are talking about when they're talking about misinformation. They
Adam:
Right.
Mackenzie:
are really convincing. They look like they might actually be there.
Adam:
Yeah, you could see this sitting inside of an article that says, could you believe that Hobby Lobby did this? And then the right type of person going, I can't at all. This is my whole personality now.
Mackenzie:
Mm-mm.
Adam:
Right?
Mackenzie:
So a lot of people are going to have to pick up the personality trait of like floating through life wondering if they're being punked 24-7, which I think most people just don't have the background to execute that skill very well. So marking these things is going to go a long way.
Adam:
It's probably fair. I mean, that's kind of what we're after. We're after when we talk about reading comprehension and media
Mackenzie:
Yeah.
Adam:
literacy. Um, it's just gonna, it's about to explode in a way that's very, very interesting because it, it breaks the brains of people who grew up with the idea that you could trust a camera.
Mackenzie:
Well, yeah, ideally, most people not being media literate is a good way to have social cohesion. So just give them, what are they called, in bowling, the bumper rails. Don't
Adam:
Hmm.
Mackenzie:
make it too complicated.
Adam:
Yeah, exactly. Okay. Uh, before we jump into today's topics, a reminder to like and subscribe wherever you're watching slash listening, throwing a comment, write a review. Uh, these all, these all help us reach you again, but also help us reach more people in the algorithm so we can, so we can make this a more of a two-way conversation. The reason this ended up being a daily show is because everything's happening so fast that it wasn't worth it to do a roundup at the end of the week. Similarly, we need to know what questions, uh, you all have as listeners and people engaging with the show. So we can, so we can keep diving into that stuff and we can all accelerate forward, get better at building with AI, things like that. Okay, let's jump into the topics.
Mackenzie:
Yeah, Josh is looking forward to the Vision Pro, he said.
Adam:
Yeah, we're supposed to lead with that. I put the tab in the wrong place. Hi Steve. Ooh, I messed up even worse than that. No, I didn't. I just got confused because it's a thread reader and not just Twitter. Smart. Smart passed me. Okay. Uh, first up, yeah, this is a, this is a tweet, so it doesn't really have a headline. The one I wrote is can Apple's vision pro read your mind? This is a tweet thread from a fellow named Sterling Crispin who, uh, I have at least verified that he had these jobs. Otherwise, it's not someone that I know. This just has been making the rounds. He's a scientist who worked up until a few years ago on aspects of the interface layer for the Vision Pro. We talked specifically about some of the specs yesterday and the compute power that actually lives on the device. So check out yesterday's episode for some more of that stuff and some links to some roundups that are kind of, Hey, here's what's happening. This one, I wanted to jump right in the weeds of, okay, it gets weird. As soon as you have that sort of compute power and these, this, this sort of, uh, this interface device where they can tell where your eyes are moving. Uh, and that's what this thread is about the work that they did internally around these, these input mechanisms and stuff like that.
Mackenzie:
So specifically, they've built, it looks like a ML model that predicts clicks because the Vision Pro is meant to be peripheral-less, no mice, Look-Mod, no mice, Look-Mod, no keyboard, okay? So one of the ways that they figure out if you wanna click on something is by tracking pupil dilation because when you click on an icon on a computer, turns out your pupil dilates at the moment of click because you're expecting the screen to change. So it gets a little bit bigger. So you can take in the new information that you expect coming your way. So part of indicating that you're clicking is that your people dilate. So they watch for that. And then because the display has a 12 millisecond refresh rate, before you could blink in like half the time of a human blink, after your pupil starts dilating, that program is open. There's other gestures to confirm that that's what's happening, but they start like preloading the new program based on eye activity.
Adam:
This, this, I wanted to call this one out because it's right at that edge of like the reason that it exists as a tweet thread and gets traction over there is because it's, it's that level of crazy, but it's for sure real. Uh, this, I haven't seen a lot of like using eye motion for this kind of stuff, but I've read studies before on. Like they can strap something to your throat, electrodes, essentially to your, to your throat. And then pick up the electrical signal intended to go to your vocal chords. And then you can think about words and move a mouse around the world. Um, so this is for sure real. And,
Mackenzie:
as
Adam:
and
Mackenzie:
another.
Adam:
if nothing else, it lets me use a crazy ass headline about mind reading.
Mackenzie:
Hehehehe
Adam:
But like, we're kind of at that point where there's, there's a lot of interesting conversations that with this VR stuff, but also, like you said, ML models that are at play when you're talking about Um, how do your eyes behave and like weird things, right? Like I can see how this would get to a health app that could say, look, we're strapped to your face, a good percentage of the day, we're a hundred percent sure you have ADD because of how your eyes behave.
Mackenzie:
Well, yeah, let's get into the other, let's get into the other pattern, which tin foil hatters are going to love. They want to know your mood. They want to know like your cognitive state. And these are for things that are good, like understanding if you're relaxed, if you're trying to use a relaxation app, or if you're excited, if you're doing something exciting like a video game, or if you're learning, if you're using an education app. And the way that they decided to track that information is subliminal images. to check congruency, they'll show you relaxing images. And if your reaction to the subliminal, like it's happening so fast that you can't detect it, it doesn't interrupt the experience, your reaction to the subliminal imagery shows congruency with clip tags on that imagery. So by MK altering you, right?
Adam:
Right.
Mackenzie:
Like showing you literally subliminal messaging, they're able to determine what kind of brain activity you're having. And based on that data, they're able to change the environment that you're experiencing in VR to promote that state that you're desiring, they've explicitly desired further. There's no way, there's no way that they're gonna like gear up the fear knob when you're reading news. You know what I mean? There's no, like it's only ever gonna be user input. There's no way that they're ever gonna give
Adam:
Interesting.
Mackenzie:
you different emotions than what you asked for.
Adam:
That's both crazy and like I said, gets that black mirror place, which is why I threw it out. But at the same time, I don't mind if I can take it off. As device makers go, what Apple has done so far really effectively is telegraph that they care about giving me that dial. And I can turn that dial to like, hey, just stop feeding me the news that you can tell from my eye behavior makes me unhappy. Cause man, when I try to do it like consciously, I am bad at not checking Twitter for the outrage machine
Mackenzie:
Yeah.
Adam:
that in the end is like probably why I have cortisol and like heart problems.
Mackenzie:
Yeah, a lot of those questions
Adam:
Uh.
Mackenzie:
that were coming up at Congress about like eye tracking weren't true about the iPhone, but they're going to be true about this.
Adam:
Yeah, they have to be. That's the interface layer. Like on some level they have to do this. Anyway, vision pro big deal. Um, Josh says, I always had an idea for VR product that worked while you sleep. I wonder if mind reading is the way there. Um, I used like a sleep tracking headband for a while that had that kind of stuff on it, where again, you can get input. Like they can read your brain waves while you're asleep. The thing that I think is harder is. Imagery is a really good way to put data into your brain through your eyes, but if your eyes are closed We're not so good at beaming things into your brain while you sleep yet But that's the kind of thing that Neuralink is trying to solve if you want
Mackenzie:
Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe
Adam:
to go get a chip put in your brain Okay, oh moving on
Mackenzie:
You swapped on me here. This is a new story. I was prepared to talk about
Adam:
Oh,
Mackenzie:
in videos, 51 terabyte ethernet switch, but
Adam:
yeah,
Mackenzie:
this
Adam:
we'll
Mackenzie:
is something
Adam:
talk about
Mackenzie:
different.
Adam:
that tomorrow. This this drop today. So or get late last night. Bloomberg reporting chat GPT maker open AI is staying private so it can make quote strange decisions. Yes, this is Sam Altman. has said that he's not interested in taking the startup public because he wants to maintain full control over the technology as it becomes more powerful. I can see the argument for this, but at the same time, it flies in the face of some of the stuff he has to say on podcasts around the idea that like government should have done this. Like one of the things that he has said that I appreciate is, uh, formerly projects this big were done by governments, the Manhattan project for the nuclear bomb, NASA getting people to the moon. Like those are public by nature products. which means a level of oversight. The problem with the private public thing when you're talking about companies and this kind of stuff is like, what he's saying here is he doesn't want to deal with the inherent oversight where you go public. Going public means you are allowed to raise a certain amount of money from your average person, but you subject yourself to a certain level of oversight, or at least transparency.
Mackenzie:
He has, he would have additional responsibilities that might conflict with things that he has to do. So
Adam:
Which is fair.
Mackenzie:
just exploring,
Adam:
Like I don't want to deal with that either.
Mackenzie:
exploring the space here, um, something, so an example of something that might seem strange to an investor is turning the company off and saying, Hey guys, we can't solve alignment. Let's stop. You can't do that. If you're public, you would never ever, ever be able to do that. You would have to get your hands away from
Adam:
Yeah,
Mackenzie:
the abort
Adam:
you'd
Mackenzie:
button
Adam:
get sued
Mackenzie:
forever.
Adam:
by the shareholders,
Mackenzie:
Yeah.
Adam:
you know, for breaching your fiduciary duty to make
Mackenzie:
Yeah,
Adam:
more money.
Mackenzie:
even if it's the right thing to do.
Adam:
And so there are very real reasons to do this or to talk about it this way. At the same time, it's this, it feels like this constant back and forth as you watch the leaders of these companies try to try to reconcile the aspects of this, right? Um, like what you're talking about is a very real AI safety reason to maintain a structure where the switch can be flipped off faster.
Mackenzie:
Mm-hmm.
Adam:
if we have rampant unsafe acceleration.
Mackenzie:
And it's not like you can't get exposure to this market. Like,
Adam:
Google
Mackenzie:
you
Adam:
and
Mackenzie:
know.
Adam:
Microsoft, Google is a public company.
Mackenzie:
Yeah.
Adam:
So they're going to have this conflict of interest where some shareholders might stand up and go, yes, but you could be making X amount more if you did this thing that might destroy humanity. Um, and your fiduciary duty, at least in the U S corporate sort of model and the way that the courts have treated it most recently is just to make money for the shareholders.
Mackenzie:
I'm throwing a flag on that one, Adam. I'm curious because I know that Google restructured from just like Google search to Alphabet and now they have like some private companies under their purview. And I wonder if like the operations are these private companies for
Adam:
Good
Mackenzie:
similar
Adam:
point.
Mackenzie:
reasons.
Adam:
Yeah, you could structure it in ways that would protect you from that. But in terms of, I thought you were going to throw a flag on the legal side, which is like, no, that is the regime currently.
Mackenzie:
Mm-hmm.
Adam:
I think it's wrong from a legal scholar standpoint, but it's where we stand in terms of the decisions of the Supreme Court.
Mackenzie:
Yeah.
Adam:
their take on the fiduciary duty to the shareholders.
Mackenzie:
I don't want to be like recommending any kind of financial stuff, but just to elucidate the market. If you wanted exposure, if I wanted exposure, this is what I would do. I would go with Nvidia and probably Facebook because the Facebook science team is really good at implementations. They invented tool formers and they're doing multimodal. So sad, so sad that you can't buy OpenAI. That's what I would prefer to buy, but these are the other. It's not impossible to get exposed.
Adam:
And I do understand all of it's take on this aspect of it. I just love that they said strange.
Mackenzie:
Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe
Adam:
Yeah. Get strange. Okay. Moving on next up, we got another tweet thread that lays out a bunch of stuff. Cause it's good summary, but you find this anywhere. If you Google or follow the links in the tweets, uh, the lead tweet says Google just dropped a 100% free learning path on generative AI with nine different courses. Intro to Gen.AI, intro to LLMs, intro to responsible AI, intro to image generation, encoder decoder, attention mechanism transformers, and BERT image captioning, Gen.AI Studio. Um.
Mackenzie:
Those are the courses. There's an additional quest at the end. I'm not
Adam:
There's
Mackenzie:
sure
Adam:
the courses.
Mackenzie:
quite what a quest is, but I think that it's like a project-based learning.
Adam:
Yeah. Yeah, this one goes with our stuff from yesterday, chat GPTs or open AIs, you know, tips on how to use chat GPT. It's exciting to me to see this back and forth. There's this new emergent thing. It behooves them to help you understand how to use the tools. And so they are going to be publishing things to help use the tools. Um, the other cool thing here is Google. I think this aligns to Google's like certification learning path stuff. So if you do enough of this stuff, you'll end up with a cert that Google will recognize. in lieu of a diploma in many cases, if you're trying to work for Google.
Mackenzie:
Josh says he's got to check it out to brush up his Gen. AI skills. I think that it's so new that everyone has this feeling of like, I don't know enough. And so when these things come up, there's, I don't know, as a software developer, there's something that we're all familiar with in our industry called Tutorial Hell, where just one more tutorial, one more example, one more demo project will get me there. It'll break me into the industry. And it never really does. So yeah, it's worth it to brush up on your skills, but start building. before it's too late, right? Like take the, don't spend all this time, like just on courses, start making stuff as soon
Adam:
Absolutely.
Mackenzie:
as you can.
Adam:
Yeah. I had a funny, I was, uh, searching for a CTL at one point and I brought in a friend of mine and as we were going through the process, I said, look, I've worked with you, I trust you, but I got to run everybody through the process. We're going to do this test. Uh, and the guy's answer was like, look, we're going to do a co I said, we're going to do a coding test. His answer was, Hey, I don't mind you watching me Google stuff.
Mackenzie:
Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe
Adam:
And like, to be completely honest, that's the right answer. In some contexts, when you're talking about building an app, especially at the time as crypto company, so like, I know he's not going to know how to do a bunch of the smart contracts things yet. There's a very rarefied set of people that knew how to do that four or five years ago. It's the right answer, right? You just, you just, you have to start building and just knowing how to do stuff. And like I said, the reason this project, this podcast is daily is because like, I don't know either. It was partially a forcing function for me to learn the same stuff. So stick around if you want to. Keep working on that stuff with us. And
Mackenzie:
That
Adam:
yeah,
Mackenzie:
brings
Adam:
check
Mackenzie:
us.
Adam:
out, check out these courses, which brings us to a
Mackenzie:
the workshop.
Adam:
smooth transition
Mackenzie:
Because
Adam:
to
Mackenzie:
we're doing the same
Adam:
our
Mackenzie:
thing.
Adam:
workshop.
Mackenzie:
We're like, here's tutorials from us. You should watch our podcast to get our tutorials. We're going to spend the next while going through the guide from OpenAI on maximizing your effectiveness using those systems. And today, we're going to talk about identity. So Adam, did you have any preconceptions or questions about this before we dove in?
Adam:
We covered it a little bit previously. We call it role playing, but I like the idea of redoing it, talking about personas, cause that's sort of the way that, that open AI is looking at it. So it means the places where you plug in, they're going to be asking about identities and personas and not. What do you, would you like me to role play as? Um, but yeah, to, to maybe start from just reading the snippet from the documentation as a tactic, it says, ask the model to adopt a persona. System message can be used to specify the persona used by the model in its replies. You pulled up a bunch of hilarious examples of this, but this goes to a thing that it's a, it's an interesting. Shift in how we think about using these tools, because if we are used to search engines, you previously would just type in your question and then from there, they would figure it out here. You're getting into that space where what we're used to is society through jobs and roles, handling the part where an agent is involved with this. And so you don't go to a lawyer and then say, okay, I need to answer. I need you to answer this question as a lawyer. But if you're talking to a generalized AI, you need to be that specific because it's, it's not going to know how to answer your question. If you don't say you're the world's best trial lawyer, write me an opening statement for the, these details of the case. If you just ask it to do that, you're going to get this weird conglomeration of things that don't fit. An actual opening statement and are instead made of like movie scripts and, you know, generalized knowledge and an effort to answer your question. But you can also
Mackenzie:
Yeah.
Adam:
get weird with it, which is the stuff that, that you've got screenshots for. So.
Mackenzie:
Yeah. So I don't know how you want to share those. I guess they're going into show notes for the podcast
Adam:
Someday
Mackenzie:
and
Adam:
we'll put them in show notes.
Mackenzie:
yeah.
Adam:
In the meantime, I will just pull them up in the links window, because I can do
Mackenzie:
sweet.
Adam:
that out of Discord, right? Make sure I got the right things fired up here. In the meantime, this is the kind of stuff that gets cut out of the podcast. What do you think, Josh? How we doing? Okay, how the hell do I open this? Copy link, open link, there we go. bottom to top. Okay. I think this gets us everything. Is this the first one? It is. Okay. So we got to do word pictures on this one for the podcast listeners, but for the people in the show notes or the live stream, we actually have the screenshots from, from Mac, um, in GPT boss or chat GPT, I guess running, running through persona work. Um,
Mackenzie:
Mm-hmm.
Adam:
anything to say about it broadly first though,
Mackenzie:
Well,
Adam:
Mac?
Mackenzie:
I'm watching the screen stream and it looks like the share screen is delayed. It's back on the Vision Pro
Adam:
Whoops,
Mackenzie:
tweet thread.
Adam:
I forgot to hit the transition button in OBS. Give it a sec. I forgot you can't see what I'm sharing now that we switched out of the old way of doing it.
Mackenzie:
I got it pulled up too. I just wanted to make sure that it would be there.
Adam:
Okay, no, we're good. Could you
Mackenzie:
Yeah.
Adam:
summarize this page is the one I'm looking at now.
Mackenzie:
Yeah, so this is a chat, this is from GPT Boss, which is a site that I created. And the reason that I like that app that I made is just the idea of like prompt management. The thing that it fixes for you is prompt management for business activities. So if you haven't heard of it yet, there's like 74 AI employees. And when people hear AI employee, they're like, oh, it automates work for me. So I could just assign work like in a ops management way and then it gets done. It's not really like that. It's just different identities that help you understand the different roles and then can also do some of that work for you. So this screenshot is talking to the copywriter, right? There's one copywriter in the website. And what that does when we say, you are a copywriter, you work for my company, this is what my company does, setting that in the system message accesses a different latent space. So when we're running through some of these examples, every time that he's doing it, he's doing it just as a professional copywriter. So there's a couple of connotations around that, like professional copywriter. It's kind of like a dry, boring, you know what I mean? Like you just got to be straight ahead, straight arrow corporate kind of drone vibe. And that's what we see in this example. There's another feature that we're showing here, which is like link reading, but it's to a link on my website. So, you know, don't take my word for it. Try it out yourself eventually. But we could see that like, boom, the link goes in and then there's two like system messages that happen every time you pass a link. But then the third like paragraph. starts the response from GPT-4. So could you summarize this page? Turns into, yeah, absolutely. And then there's this kind of dry professional summary of what this service is, including pricing and follow-up and how there's a fact section. And you could get some of that information. If we could go to the next slide to start showing this working in a different way. Adam, confirm if you've done that, because I still can't see
Adam:
Yes,
Mackenzie:
it.
Adam:
done.
Mackenzie:
OK, cool. So. The next slide is, could you summarize this page in a language that Dan Kennedy would use? If you're not familiar with Dan Kennedy, he's kind of like a carnival barker copywriter. So we see when this stuff comes in, the language is a lot different, but it still hits all of the like professional requirements of being a copywriter, right? This is summarized as a copywriter would, but it's summarized as a copywriter who's channeling spiritually Dan Kennedy. And so. I think that this is a really good strategy. I think that like GPT boss or chat GPT users should be doing this. I love Dan Kennedy. I think that his stuff is engaging and it just makes the most sense in my mind. I think he's the best copywriter ever. Um, so if you had to pick one guy to channel from the dead forever and ever, Dan Kennedy would be a good one. But as we continue on, I have some more creative examples of the same kind of activity.
Adam:
for people not looking at the screenshots. The first one says introducing just-in-time content. The next one says discover the ultimate content solution.
Mackenzie:
Yeah.
Adam:
Right, like there's a difference here. And Personas is a more stylized and complete way to jump straight to, no, I needed to have this vibe if you don't necessarily even have the language to say, no, I need it more excited. I need more emojis, I need blah blah.
Mackenzie:
Yeah, exactly. So, yeah, in the previous paragraph, it says this is a product that solves issues with long sales cycle times. And then in the Dan Kennedy one, just time content, the breakthrough service from GBT boss designed to give your sales marketing efforts the rocket fuel they need to skyrocket your results and trim those sales cycles down hourglass emoji chart increasing in value emoji. And
Adam:
Right.
Mackenzie:
then a couple more examples, there's Dawn Draper, which like pulls it back. Of course, this is a fictional character and it's very like. 50s showbiz baby kind of vibe. And then the final one, I don't think I sent this one to you, Kurt, but I did it again with Patrick Star from SpongeBob SquarePants.
Adam:
We got,
Mackenzie:
Do you have that
Adam:
we got
Mackenzie:
one?
Adam:
Don Draper,
Mackenzie:
Yeah.
Adam:
which says introducing GPT bosses, just in time content, a game changing solution to elevate your brand's presence. I mean, I don't want
Mackenzie:
Yeah,
Adam:
to hear the
Mackenzie:
so
Adam:
pitch
Mackenzie:
it's still,
Adam:
in the room
Mackenzie:
it's like,
Adam:
for that,
Mackenzie:
it's hype.
Adam:
you know, spark a cigarette, it's hype, but it's way more. How do we
Mackenzie:
Refined,
Adam:
explain
Mackenzie:
sophisticated.
Adam:
how this is going to change? It's right.
Mackenzie:
Yeah,
Adam:
Exactly.
Mackenzie:
yeah. So
Adam:
Okay.
Mackenzie:
that's identity priming through chat. You're able to select a celebrity, fictional or otherwise. And because that's in the latent space, the agent has access to that. But because we're using the copywriter, we're not missing things. Like if we were using an AI, the AI would be like, I'm just an AI. I don't know about human psychology. I don't need to do this properly. It can just do... this is what Don Draper would say. And maybe it would use the wrong parts of Don Draper's personality to get this done.
Adam:
So. Uh,
Mackenzie:
Sorry, go ahead.
Adam:
and you're doing this essentially through prompt injection, right? Like, like GV boss adds some stuff into the flow that makes sure that it comes out this right way, because it knows that you've selected this tool, you want a subset of things that like there's persona-ness in there also that
Mackenzie:
Mm-hmm.
Adam:
is layered on top of the stuff you're talking about here.
Mackenzie:
Yeah, so rather than an AI pretending to be a copywriter, GPT-BOSS goes one step further. It's an AI that is a copywriter that's pretending to be Don Draper.
Adam:
Right.
Mackenzie:
And then.
Adam:
And then the next one you've got, you start running through some, we've got third grade, fifth grade, and then
Mackenzie:
Yeah.
Adam:
me troubleshooting some Steam Deck. Thanks. Third grade. There we go.
Mackenzie:
So the tools, the content spinner for third grade, fifth grade and the luxury brand, I'm not sure if I sent everything in, but these are examples of using this programmatically. So I added the tools to GPT boss for when somebody doesn't want to do this kind of prompt engineering. They don't want to type out the word, could you rewrite this into a third grade reading level because that does
Adam:
Right.
Mackenzie:
take some time. They could just click a button and then paste the content that they want rewritten. And... So an example of this is that Don Draper thing that we were just looking at. Let me compare. The Don Draper introduction was introducing GPT Boss's Just-in-Time Content, a game-changing solution to elevate your brand's presence. And then at the third grade level, we have Meet GPT Boss's Cool New Tool. Just-in-Time Content to Make Your Brand Shine. Star emoji. So it almost sounds like a children's book, and that's kind of the idea here. The purpose, by the way, of the third grade content spinner is to identify basically logic errors. It reduces the language complexity so you can't hide behind like verbal ambiguity.
Adam:
Right. Yeah.
Mackenzie:
Yeah.
Adam:
It's not about the idea of now you're going to sell this to third graders. It's about like third grade level is a good way to flag a certain complexity of communication that for certain contexts, like that's how I would look at it as well, right? Like,
Mackenzie:
Mm-hmm.
Adam:
look, this is copy for a brochure. People are going to look at it quickly on a train. You can't be more complex than X thing. Right.
Mackenzie:
Yeah, that's.
Adam:
And, and aligning to education levels is a good way to do that. That AI
Mackenzie:
Exactly.
Adam:
is, is good at.
Mackenzie:
So
Adam:
Thanks for watching!
Mackenzie:
and then the next step, and we don't have a screenshot for this, because it doesn't make a ton of sense in a screenshot. But you if you have, for example, a manuscript, or if you have, like a backlog of blog content, say that you've been like running a blog for the past 10 years, and you're a subject matter expert in something complicated, like chiropractic accident recovery techniques, you could run through programmatically. that backlog through a prompt like this, through your fifth grade content spinner, rewrite all of your blog stuff to remove the curse of knowledge that you as the chiropractic experts suffer. And there's like, that's just one specific example but I hope that you can see the potential of this across domains and from now into the future. It could be a filter on your WordPress blogging. You could just set this up inside of WordPress so that everything that you push to publish gets rewritten into a fifth grade level so that you're not alienating ESL readers.
Adam:
Super cool. Like, like we always say at the end of this tune in tomorrow for more of this good stuff and for more on what's happening in the news. Otherwise, this has been Accelerate Daily for June 7th, 2023. I am killing time while I try to find script. There we go.
Mackenzie:
Yeah, you might want to poke at that again.
Adam:
Okay, that's Accelerate Daily for today. Like I said at the top, if you've got something out of this, like subscribe, even write a review, those metrics really do make a difference when it comes to reaching more people working on the future of AI, like we are, and, and really helps us surface questions for like the previous, you know, the previous stuff. We can, we'll retread this stuff over and over again, if, if it's helpful to understand what's going on. Uh, and if you can make the schedule work, jump in the live stream. Uh, that's one of the best ways to interact with us on this stuff. Thanks again, everyone for joining us. Uh, this has been accelerate daily. I'm Adam.
Mackenzie:
I'm McKenzie.
Adam:
Take it easy, everybody. Okay, let me do this outro read. Accelerate Daily is produced by Adam Kerplman and Mackenzie Bose and edited by Steve Koch. The opinions expressed herein do not represent the official opinion of Mission Control. Thanks again for listening everybody. We'll see you tomorrow. Okay, going to stop the record.