Meta's Bold Move: Offering LLaMA 2 for Free - An AI Model Competing with ChatGPT 3.5

0:00What can 160 years of experience teach you about the future? When it comes to protecting what matters, Pacific Life provides life insurance, retirement income, and employee benefits for people and businesses building a more confident tomorrow. Strategies rooted in strength and backed by experience. Ask a financial professional how Pacific Life can help you today. Pacific Life Insurance Company, Omaha, Nebraska, and in New York. Pacific Life and Annuity, Phoenix, Arizona. The holidays are here, and that means it's the most wonderful time of the year to save with Rakuten. Use Rakuten to stack cash back at your favorite stores on top of holiday sales.

0:40That's savings on savings. With Rakuten, you get cash back on gifts for everyone on your list. From toys for the kids, to kitchen gear for the person who loves to cook, to electronics for everyone. You can even save on something for yourself. Just shop the stores you love and Cashback is automatically added to your account. And you can get paid with gift cards, PayPal, or check. Or eligible American Express card members can even choose to earn membership rewards points instead of Cashback. It's truly a no-brainer. Join for free today and get a new member bonus after minimum qualifying purchases.

1:17Just go to Rakuten.com, download the app, or install the browser extension. That's R-A-K-U-T-E-N. Terms and conditions apply. This is massive breaking news out of Meta. I know I've been releasing a lot of Meta-related podcasts lately, and they have been on absolute fire. But this one, I believe, is one of the biggest stories. And that is that Meta has just announced, in partnership with Microsoft as their preferred supplier, that they are going to be open-sourcing the Llama 2 model. Now, for those of you that don't know, So Lama 2 is Meta's equivalent model to ChatGPT 3.5. So not GPT-4, more than GPT-3, but about 3.5.

2:00But the thing that is amazing about this is they're open sourcing it. So they're literally giving away the equivalent of ChatGPT 3.5 for free. So for any developers or businesses that are currently integrating AI, of course, they're all very familiar with the fact that they have to go and you know you got to pay for every single API call that you use to chat GPT it can get pretty expensive and the fact that they're now able to go and put you know use something like Microsoft Azure they said they're going to be on Hugging Faces and on also on AWS in the near future but the fact that you're going to be able to go to any of those platforms and run this new Llama 2 model is going to be absolutely incredible it's going to be right you're getting this whole thing for free and all you're paying for is the compute cost which is going to be very very low so they said people can start applying to get access to it and i think what's really really crazy because it doesn't stop there the news gets pretty crazy i think all of this was recently announced at microsoft's inspire event and so you know they kind of did it saying hey look we're partnering with azure they're a preferred supplier they're going to put it on all the other ones but they're saying azure is like the preferred supplier and i'm assuming that the partnership there is going to get some sort of special perks.

3:09But and then in addition, I think they do it because I don't know, maybe they decided Microsoft was their player instead of Amazon or Google, who are all doing, you know, similar have similar things. But OK, the crazy news does not stop there, because actually at the same time, Qualcomm announced that it's going to be working with Meta to bring Llama to laptops, phones and headsets starting next year. And they're going to have AI powered apps that work without relying on cloud services. So essentially, you could have an application, an AI application, like an image generator, and it's not going up to the cloud to try and generate some sort of AI image for you.

3:49It's literally doing all that processing on your phone, on your laptop, or on your VR headset, which I think is absolutely insane because that is going to be, again, another massive cost savings for any developers and any companies that are working in AI, right? Because right now the big cost that you have is these API calls to open AI. I've heard some companies paying 30 % of their operating costs are going to open AI if it's a real heavy thing. So this is going to be massive. It's not just the fact that, okay, look, we can switch and get on Microsoft Azure and get like a 10th or 100th of the price perhaps.

4:26Now it's literally going to be free. Like processing is going to be free. it's just going to be done locally on a user's device, whether that's their phone or their laptop or their headset or whatever. And that's all thanks to Qualcomm, who's, you know, specifically integrating this. Something that Meta has been absolutely amazing at so far is the fact that they're able to build these models that use way, way less processing power. Like, I can't explain to you how incredible it would be to get something like GPT 3.5 on your running locally on your phone instead of having to go over to, you know, OpenAI's data center and run just like astronomical amounts of compute through it.

5:05So this is actually a really, really big move. So what's interesting is they said Llama 2 was actually trained on 40 % more data than Llama 1, right? So they initially launched Llama 1 earlier this year, maybe back in February, if I'm not mistaken. And a few days before or after they launched it, the code for it was actually leaked to 4chan. And so people were able to essentially get the code for that. And I think at this point the cat's kind of out of the bag and meta could try to lock down this next version but realistically um they were open sourcing the last version so they open sourced this version as well and i think they're doing this at a big play to beat chat gpt they know their only hope of beating chat gpt at this point may just be to make this thing open source free everyone can use it and this is the original concept of open ai right um it's interesting hearing from sam altman he's like no like i've heard him go on a bunch of podcasts and be like hey should i open source GPT-4, what do you think?

5:59And he's like, ultimately, we're not going to do that. Okay, obviously, you got$10 billion from Microsoft is not going to do that. He needs to like kind of maintain this. And also, on the other hand, some people think it's a strange like why would Microsoft be partnering with Facebook so hard when they know this product potentially could kill their OpenAI investment. And I think the response is obvious. Microsoft is wants to be on the winning team and they want their Microsoft Azure platform to benefit. They know ultimately OpenAI is a boon for Microsoft Azure because if OpenAI is on Microsoft Azure and all of this all of this processing and compute is going through their platform it's going to be you know they're going to be the cloud provider that's doing all of that so they're going to make the benefits but if they were to you know shun someone like Llama from Emeta and be like no you're a competitor you know we don't want to do the partnership okay great Google Cloud would snap them up and then Microsoft would just lose all of that and if Llama ends up beating ChatGPT then all of a sudden Microsoft just loses out on a massive segment so I think Microsoft is really just at this point trying to woo anyone they can we've seen they put some heavy investments into inflection AI which is another player that's like chat GPT so now they have their inflection their open AI and their lama partnerships and between the three of them I think Microsoft Azure is going to be very set to see some very very steady solid growth on its cloud computing platform so I think that's going to be a massive win for Microsoft as well.

7:22In their press release, they kind of explain the decision to open Lama up as a new way to give businesses, startups, and researchers access to what they say is more AI tools. And really, I think this is something that's going to allow for a lot more experimentation just as a whole community. So according to Meta, this is obviously trained on a much larger data set. And they say that it actually includes information from publicly available online data sources. It also says that this outperforms a lot of other LLMs like Falcon and MPT when it comes to reasoning, coding, proficiency, and knowledge tests.

8:00And so I think what's really impressive here is just the fact that, you know, Meta says it wants to improve safety and transparency. And so I think that this, you know this llama 2 model has been red teamed or tested for safety by generative adversarial prompts to facilitate model fine-tuning and they've done this both internally and externally so meta really has tested this new llama 2 model for safety and of course i think this is really important if they're going to be open sourcing this thing they want to make sure that this isn't you know some diabolical ai that's going to ruin humanity which of course it's kind of left to be said if it is, but it appears that they have put a lot of resources and efforts to ensure that that doesn't happen.

8:44They specifically said in a statement, we believe an open approach is the right one for the development of today's AI models, especially those in the rapids or in the generative spaces where the technology is rapidly advancing. Opening access to today's AI models means a generation of developers and researchers can stress test them identifying and solving problems fast as a community. So I think this is really massive news. I think it's got a lot of big implications for AI in general. I think OpenAI better watch out because this is big. This is the biggest one. This is equivalent of almost like Google Bard being open source.

9:22Not Google Bard being free, but Google Bard being open source, meaning anyone could theoretically take Google Bard and run it. And eventually, this is crazy because if it's open source and Qualcomm's trying to make this thing so it can run on devices, Like you could grab this open source model and have your AI applications running on people's personal laptops and phones and headsets. Absolutely insane. Some massive, massive implement implications because ultimately what this is going to do is make it so that a lot of AI businesses that were not viable from a financial perspective or standpoint before will absolutely be viable.

9:54Right. So something that maybe doesn't make a ton of money, but it's super, super helpful is now going to be a completely viable option if it's running locally on a device. I'm super excited to see where this goes. I think this has got some massive potential, so we'll definitely be following this story very closely.