Meta's AI Power Play: Llama 3.1 405b, Reel Search

Summary

In this episode, Dalton Anderson discusses Meta's upcoming release of their 405 billion parameter model. He mentions that the model will be open source but the model weights may not be released. He also talks about Meta's new feature on their Meta AI website that allows users to search for reels. Dalton shares his experience using the feature for fashion inspiration and discusses the potential use cases for the feature. He also provides updates on the progress of his home remodeling project.

Keywords

Meta, 408 billion parameter model, open source, model weights, Meta AI, reels, fashion inspiration, use cases, home remodeling

Takeaways

Meta is set to release a 405 billion parameter model that will be open source.
Meta's new feature on their Meta AI website allows users to search for reels, providing inspiration and ideas for various topics.
The 405 billion parameter model is positioned as an enterprise-level model that will compete with other advanced models in the market.
Open source models allow for greater transparency, collaboration, and faster bug fixes.
Dalton shares updates on his home remodeling project, including electrical work, closet installations, and plumbing challenges.

Sound Bites

"Meta's exciting release: the 405 billion parameter model."
"Meta's new feature on their Meta AI website allows you to search reels."
"The 405 billion parameter model is an enterprise-level model that competes with other advanced models."

Chapters

00:00 Meta's Upcoming Release: The 405 Billion Parameter Model
03:24 The Benefits of Open Source AI Models
13:19 Exploring Meta's Reel Search Feature on Meta AI
25:13 Updates on Home Remodeling and Plumbing Scare
31:11 Preview: Live Demo of the 408 Billion Parameter Model

Creators & Guests

Host

Dalton Anderson

I like to explore and build stuff.

What is Venture Step?

Venture Step Podcast: Dive into the boundless journey of entrepreneurship and the richness of life with "Venture Step Podcast," where we unravel the essence of creating, innovating, and living freely. This show is your gateway to exploring the multifaceted world of entrepreneurship, not just as a career path but as a lifestyle that embraces life's full spectrum of experiences. Each episode of "Venture Step Podcast" invites you to explore new horizons, challenge conventional wisdom, and discover the unlimited potential within and around you.

Dalton Anderson (00:00)
Welcome to VGSTEP Podcasts where we discuss entrepreneurship, industry trends, and the occasional work review. Today we're to be discussing Meta's exciting release that doesn't happen to align up with the episode. So I can't do a live demo, but we'll be briefly discussing it. It's their 408 billion parameter model that they were releasing later this week. Hopefully it doesn't get delayed. I think there's been a lot of hype to it, so I'm pretty sure that Mark will release.

the model and it's going to be open source, which was confirmed. I think that they're not releasing the model weights. think there was some general pushback from the public like, Hey, if it's open source, let's make sure we release the model weights that and Meta released a new feature on their Meta AI website where you can search reels. And I think that there's some interesting use cases potentially people might use them for and how it might be useful to yourself and what I was using it for as well.

And then some general updates on the success of the jobs and the nightmares that I had to working on Nana's home the last couple of weeks. I had some scares and we got through them. It's a general sense of what we'll be discussing today. Before we dive in, of course, my name's Dalton Anderson. I'm your host. My background is a bit of a mix of programming, data science, insurance, offline. can find me running, building my side business or lost in a good book. If

listen to podcasts and video and audio format, YouTube is the place to be. If audio is more your thing, you can find the podcasts on Apple podcasts, Spotify, or wherever else you get your podcasts. Okay, so Lama3, 408 billion parameter model. So right now Lama3, have a 70 billion and a seven or eight billion.

They have a seven, they have an eight billion, they have an eight million parameter model, a 70 billion parameter model. And those seven and eight billion parameter models performed very well against their counterparts. And in some cases, even they were lumped in. So typically they'll classify the models like say like small, medium, large, like a large would be like, okay, this 400 billion parameter model we compared with other large models, but they don't necessarily have the same parameters.

the amount of parameters. So in certain cases like this Lama three, eight billion parameter models compared to other models with 15 billion or 20 billion parameters. And there's 7 billion parameter model that's mainly made to work on wearables like the Meta glasses or I don't, I know the Meta glasses uses their small model. They're, they're 8 billion.

but I'm unsure whether Quests, MetaQuest uses the eight or the 70. That I don't know. But basically the models perform fairly well against their peers in that same category with most of the time those respective MetaModels having less parameters than their counterparts, which is pretty cool. And I was pretty impressed.

I wasn't pretty impressed. I was impressed with how the models performed with the things I asked it. It seemed generally pretty legit for only 70 billion parameters because when you go to meta .com, you're only using the 70 billion parameter model because the 408 hasn't been released yet. But what is really cool about Meta's approach and similar to X .com

I guess otherwise known as Twitter, old Twitter, or X. It's called X now.

X and Meta have taken the approach of putting large sums of money to train these AI models and then open sourcing them. And so if you go to meta .com, there's a big button in bold to download the models. All you got to do is fill out your name, like who you are and an email and you can download the models. it. You just got to register yourself and they allow you to download the models. Apparently with the leaks that they're saying,

There was some leaks of the 408 billion parameter model and it was a download payload of I think 800 gigs or something like that. So it's almost terabyte of stuff that you have to download to run the model. And that's not even to run the model, that's to the model on your machine for you to run it. You probably need to either combine the resources of multiple computers.

or have a really powerful GPU like rig or something like that to use for compute for the model. I don't know. I haven't really looked into it that much, but I do know that that's too much to run on a computer. One computer at least you could maybe combine two really good computers and be able to run a model, but not to its full capabilities, obviously.

That being said, you could, you could pick up and download and make your own little app for free.

The 70 billion or the 8 billion parameter model, like that's not that much like 8 billion parameter models like may just be stored on watches

glasses and stuff like that. And that's what they, I know for a fact, that's what Meta uses for their, what are they called? Meta Ray -Ban glasses? That's the model that they use. And people have said that they find it useful and eventually they're gonna get it to the point of being able to see, like you also give it permission, like, hey, a Meta, can you look at what I'm looking at?

Can you tell me what year this car was produced or something like that? Maybe you're looking at an antique car or something. I don't know, good example, but, or how to make this. And maybe you ask it how to make a certain type of dish. Like what ingredients does the dish typically have? Or maybe you don't even know the dish and you're like, what is this dish? And maybe it finds out for you. But at one point you'll be able to,

talk to your Meta and be like, hey, when's this place open up? And then without going to your phone, Meta will search the internet, not the internet, because it's Meta glasses. This is a different situation. Eventually they would want it to search the internet, but I don't know where that's gonna be with the glasses. Maybe it's connected to your phone via Bluetooth and somehow the Bluetooth corresponds. It gives commands to your phone to search back.

gives MetaGlass back the information and then Meta talks back to you. But that might take too long and you might as well just look it up on your phone.

Maybe not for that kind of stuff that is super duper dynamic, but things that might be set in stone for a while might be better. But basically long story short, eventually Metaglasses will allow you to search for stuff on your face without you having to take out your phone, which is nice if you're down for that. I don't know. I think it's pretty cool. I don't know how everyone else.

thinks about it, but I find it pretty interesting. So I would definitely do it. I would buy the glasses, but before I buy the glasses, I would probably buy of Meta, what is it? Meta Quest or yeah, Meta Quest 3. I would buy that before I bought the glasses, just so I could give cool demos. I could talk about it in the podcast and who knows, maybe I could just start writing this stuff off. know, I needed to talk about the podcast. Like why not?

Why not make a business expense out of it? But anyways, so the 408 billion parameter model is more of

enterprise level model. This model would compete with the ChatTPT 4 .0, Jetstee 4 .0, the Gemini Advanced, or the Pro, like 1 .5. And so that's where the 4 .0, or 408 billion parameter model is positioned when it gets released. So it's supposed to be an open source,

competition or an open source model that competes with open AI and Google and anthropic with their cloud.

Will it do that? I don't know. From the leaks, it seems like it's going to be very good. I'm not completely sure. I mean, there's a lot of hype in AI and sometimes I've gotten caught. I've gotten caught, especially with Google's AI announcement where they staged it. And there's been some, some staging, like there's been that AI, what Devon,

It's like this AI code, code bot that is supposed to be able to be super good at coding and replace a whole bunch of developers. They staged that demo. and I staged their demo. Yeah, I don't know. It's tough. That's tough. So I don't really trust things until I look at it myself nowadays, unless I can get my hands on it or there's been some independent reviews of these releases.

very hesitant to say so. But we'll see soon, because hopefully it's going to be released in the coming days. They didn't give an exact date, I don't think. They said next week, which was last week. So it's got to be sometimes this week, right? Am I doing the math right?

The potential of it is you allow researchers or students at a university research groups to play around with these models for free. And it gives people an idea of

an idea playground, I would say, that's how I'd phrase it. You have the ability to play around with the model, both models or all three for free and you get to be able to see the weights, you get to see how it was constructed, you get to see everything, which is not something you get to do with closed source models. And I think the general consensus is, hey, open source things are better.

There's a lot more good people in the world than there are bad people. And if you open source things, there's a lot more eyes and people can see what's going on in real

with security defects like security leaks or, just different defects of, of the code base can be fixed a lot faster because there's a lot more people seeing it. It's a lot of people using it. And so then when you have things open source, it just makes the maintenance of the code base a little bit easier because the public has a interest in it. All right. Not really the general public, but the coding public, I guess.

the active developer group, the community. So it's just overall better. so, you know, open source allows things to move quicker. People can provide feedback, ideas. And so hopefully that gives it a little bit more velocity. We'll see. For, can use Llama, the license, I think the license is not for everything except non -commercial use.

You can't start a business for it, but you can. And I think that it hasn't been finalized. think I think it's going to be more relatable when the four hundred and eight billion parameter model comes out because the first two are great, but they're not. my goodness. So once we get to that level, I think they'll provide some clear guidelines. know that Mark was talking about like, hey, like I don't want.

like a large massive company to be using this model for free. Like if they're big, like I'm going to make them pay. But if like some small shop is trying to get started and want to use our model to build their product, then we could let them do that for, you know, if their revenue is under like 10 million or something like that. So I think it depends on the size of the company, whether it's free or not. I think that's the structure of the license. But those things are subject to change all the time. So you have to be careful.

Okay. So that demo is coming next week. Reel search. So with Meta AI, if you go to that Meta AI website, you can now search for reels. And so I was using it at first, I was using it for fashion. I got to get my fashion up. I may have a lot of things going

for myself that are positive.

One thing I don't really have going on is some fashion. So I think my fashion sense and my ability to properly structure outfits with a defined stylistic view is poor. So I'm trying to become more fashionable, which is a difficult thing when you're starting from zero.

I guess, because you don't really know who you are, what you like. So you had to go through this whole thing where you like have to look up different type of styles and you have to look up, okay, all of your styles that you like, what is the defining color palette of that style? And then you have to figure out, okay, well, if I like this style, I might like this style. You have to just go through all these different styles and figure out, if I combine all of these different styles, can I make my personal style?

Cause you don't want to just copy, copy paste other people's stuff. You want to make your own identity. And so you could use Pinterest to do that, or you could use, or you could use this Met AI where you could search for the style that you want, which is pretty cool. I think I still like Pinterest more because you can save it, but you can get just general, general reels and stuff from this.

this searchability. Or you could ask it, okay, like what are some home decorating ideas? And then bam, then it pulls up reels, popular reels, and then you can get there. From there, you can find the suggested posts that are related to that post. And so it might find you an area where you might want to start, but you don't necessarily know where to find that first reel or that first video. And

You could search via Meta AI like, I want to see stuff relating. I said, like, I want to see stuff relating to home decor. And it shows you reels. And let me see this. I share my share this tab here. Screen.

this one.

So I am sharing this tab with everyone.

So basically if you're not watching on video, I am sharing my screen and I'm sharing of this Meta AI search feature. So I asked Meta AI, can you show me, I have some major typos in here, but it says there should be reels of, I'm not sure. There should be reels of creatives for products in the home space. So a creative is just like a video ad. And I,

It sent me some stuff related to different ideas like decor inspiration, home organization, smart home, tech, whatever. And then I said, please show me reels. And so it shows me reels of people cooking with stuff, doing some organization things. all of these reels, they might not be exactly ads, but they are advertising a product

just like this like sneaky, sneaky advertising with their creatives. And then I said, show me reels for couples. And I got one for like shopping. This was less successful. I got one for shopping. Like, I know this family bought a Dyson, I don't know, hair curler or something like that. And the husband was complaining. The wife was like, it's not that expensive. He looks it up. It's like, oh, it's like 600 bucks.

And then she's like, babe, I got it on sale. I got a deal. And then he's like, how? And then she's like, this is the website, you know? So it was a cool little story that they sold. I looked at some cooking stuff that wasn't as successful. And then I looked at some more cooking stuff that was more successful. And then I did ads for couples. And the ads for couples are pretty good to where

It is showing you things that you want to buy for your partner or you want to buy for yourself to enhance your experience with your partner. It seemed like most of the time that these reels were related in

sexual manner and it was more related to empathy, being intimate with your partner. So it was all about sex and toys and different things that, I didn't mean to click on that, different things that you could be doing for your partner, I guess. That's what I was seeing and hopefully no one's dogging me in the comments about this, like, this is just your algorithm, because it's not, so I'm not logged in.

Like it just doesn't know I'm logged in. So it doesn't know my algorithm unless that and that'd be crazy. but I don't see that kind of stuff on my reels on my, my Instagram. My stuff's more like gym motivation, running and motive, like motivational quotes and like books and stuff. So I don't get much of that kind of content, but I do watch reels and make sure I keep it in

the fitness area, it can get pretty bad pretty quick, depending on what you like or watch.

But I think that when...

the models become more advanced. when I keep, when I'm saying more advanced, I'm not talking about that far away. Like maybe this, this four or 108 billion parameter model, when I ask it these things, it knows exactly what I'm looking for. Scours the comments, scours the posts on the Instagram knows like how one thing is an ad, the other one's not an ad, but really that is an ad, even though it's not a paid ad. It's like, it's just organic marketing. And

it populates that stuff to me because that would save some time. I don't think it would save time on product discovery. If you want to do organic product discovery and just scroll through, make a burner account and scroll through and find products that would sell to your demographic that you want to sell to and help build a brand. That's an option to start out. But then you're like, okay, I want to

I don't want us to copy creatives from my product space. I want to do something completely different. And so you could ask Meta, hey, can you provide me creatives that are in this space that are labeled as ads or seem like they're ads? And then Meta would just be like, yeah, sure thing. I can do that for you. And then you can just watch all these

these ads and say, okay, like, I like what these people did. Like the way that they presented this is very nice.

There's a lot of motion going on. It's clear what this product does for me. It makes me want to use it. Let's see this. Share this tab instead. You should see this reel. This reel is really not that good, but it's got 60 likes. It's literally just them washing glass with this little thing that they put on the side of their sink. You click it and it washes

It's pretty bad, honestly. I'm not a creative expert, but I could definitely tell you that that reel is not good. Okay, share this tab instead. This is not an ad, this is just a recipe.

Let's just get one

Okay, let's do this one. Share this tab instead. Okay, link in the comments. So she's got eight likes in a comment. She's selling

I guess detachable cookware where you could attach the handle, put it on a large pot, small pot, a

Really odd video angle though.

Anyways, so you can use this feature to save you time on discovering what kind of creatives you want. Or you can see like, you can populate yourself with good or bad ads and you can see what people are doing and what's selling, what's not selling versus you having to figure that out. Or it makes it just a little bit easier to figure out what creatives you want, get new ideas. These things

forever changing. So I know less about it because I haven't been doing it in a bit. But yeah. I think it's pretty cool. And going to be useful not only just for product discovery, but for other things as well. I mentioned like a personal personal example, like to get my fashion better. Those things you can also use it for maybe a birthday ideas

you know, places to travel, those kinds of things. And you could see actual videos of where people are going, what they're doing, best restaurants in Tampa. If I did that best.

places to eat in Florida. Rails.

And let's see what they say. I will say that to search reels, to query reels, it does take a while. Like it might take 40 seconds to pull back information that you requested. All right. So I just said best place to eat in Florida. It's got stuff related to Miami, similar places. But it's just general, I mean, you could do like where you live, but I just did Florida because it's easier.

Because then if I do Where I Live, I was thinking about it, I was like, I'm gonna get too involved and I'm just gonna watch them. So I could look that up later.

But it gives an ability for people to find content that they're really searching for instead of scouring for it. And so maybe they don't have to watch as much stuff anymore where they could just find exactly what they want. Cause I don't need to scroll through 12 reels or 12 TikToks to get to what I need to. All I need to do is find the one that I like.

You know, I'm looking at this is, oh, I didn't mean it. This is one, two, three, four, five, six, seven, eight, nine, I think 10 reels per query. So you could look through and say, okay, like I don't want to look at that one. I don't want to look at that one. So I think the implications for creatives is like, or creators, you need to have, in the future, you're going to need to have a really good selling.

frame, like your, your, your starting frame for your Tik TOK or real needs to be good. If people start using this feature, because a lot of them, I'm just not going to click on them. I'm not even going to see it because it doesn't look interesting. All right. So house update. I am almost finished with helping remodel and it's home. So we,

I had some help from someone to work on the house and he did the electrical, which I don't know how to do. I did the closets, which sadly aren't fully done yet, but they're getting there. They're so close. I just need to do a little, little tweaking to mine and finish the little baby closets. They're not baby. Like they're not for babies. They're just smaller closets for the corner units. Just got to put up the, the rails and stuff for hanging clothes

other things and shelving. But the other closets are done. Cause I did the pantry closet, hallway closet, Nana's closet, my closet. So I did in total four closets. I feel like that's a good amount in a week. It takes a long time. So the guy worked on the house finished the high hats, which makes me glow. I'm glowing in this podcast video because there's so much light now

He also helped fix my attic. My attic didn't have as much insulation as it needed. And then there were no vans. I said vans. I meant safe vans. There were no gable vents. that we did have gable vents, but for some reason someone closed them. And basically the gable is when the two slants on a house meet and it makes a little triangle.

there should normally be like a vent in that area to allow, there should be more than that, but just rule of thumb, you should at least have two, hopefully, especially if you live in Florida, to allow the hot air to release from your attic. And basically, know, hot air rises, it tries to find cold air, cold air tries to find hot air. So the hot air rises in the attic, goes through the vent, the vent then circulates the cold air in passively.

And it cools down the house of the attic, the attic of the house.

We, for some reason, had someone close our vents and then cover them up. So Lewis helped cut out the framing of the house. Cut out a hole through the siding, like a square. Picture frame it, put a vent out there, waterproof it, all that stuff. Which is great

me because the house is a lot cooler than it normally is. So it's awesome. So I'm loving life, but we did have a scare. We had a plumbing scare. basically we were replacing our trim for the house. so the trim is like, I guess like the shower trim or the shower set is basically the valve, the shower valve that's inside the wall behind your toilet, not your toilet, behind your shower handle.

and then your shower arm, which is the thing that connects to your shower head and then your shower head. And then your drain overflow valve if you have a shower and tub and then your drain. If you have a shower and tub, so my bathroom is a shower and tub. So I got this drain that looked great. It matched the set that I had. And.

Lewis took out the drain, then the original drain broke upon removal, which is pretty common for older drains. Like my drain has probably been in there for 50 years or something. I'm not sure if it had been remodeled before, maybe, maybe 40 years. I don't know. It's been in there for a long time and the drain broke and my drain didn't fit. Other drains didn't fit. And so I was in this position to where if I couldn't find a drain,

I couldn't use my shower and I would have to pay for the person working on the house either to chip underneath the tub to disconnect the pipe, the drain pipe to connect a new pipe or have him rip out everything and basically redo it. Both of those things were out of scope for the project. So I was in like dire need to find a drain and luckily

I found a drain that fit my...

dimensions for my drain, which was awesome because it saved a lot of money that I didn't want to spend. And it was really cool because I was able to just put it in and just drop it right in there. mean, really not. But like you have to do plumber's putty, put it in there, tighten it down. All good. But if I couldn't have found the drain because the drain was so old, like the drain type, I was having a hard time finding the

Not the right length, but the right thread width, that the threads wouldn't fit. And so a lot of times everything was too big. And if it's too small, you can put on an attachment and basically rethread it called like a bushing, I think. But for my case, mine were too, like the new drains were too big, so I couldn't really make it smaller.

So I eventually found one, it worked out and everything is a happy, happy story. And so I'm excited, you're excited. I know you're thrilled to be hearing about that. You're like, yes, you're jumping for joy. Like, yeah, that's awesome.

Next week, we're going to be discussing the 408 billion parameter model. Hopefully we'll have a live demo of the episode, not the episode, live demo of the model during the episode, which would be great. And of course, before we end it out, have a great day, a good morning, good afternoon, wherever you are in this world, and I hope that you're listening

next week. Have a great day. Thank you for listening.