Open weight models allow companies and individuals to run, fine-tune, and operate AI models entirely on their own infrastructure, with no dependency on the original creator or big AI labs.

11:21run by Microsoft, the AI lab in China that created it has no idea that they're actually running their model. They basically handed it off and now at At point Microsoft is in full control. Now, one of the last ways you might get your food is through home cooking. You pick out your recipes, then you source all the ingredients, [music] and then when you get home, you cook it

11:41yourself. But, because you're in full control, you can improvise. You can add more ingredients, you can skip ingredients, you can change the portion size. You're completely in control. And this is one of the other things that open weight models give you. Not only can we run them ourselves or access them via some third-party provider like Microsoft, you can actually fine-tune them. That is, tweak the numbers inside of them to make it

12:04more purpose-built for what you're trying to do. And we actually see a lot of large companies or enterprises doing this. So, instead of having to be dependent on some other corporation like Microsoft, they spin up their own infrastructure internally, and then they run these open weight models, and they can fine-tune them so they can better work for their organization. You can essentially take a model and fine-tune it about specific

12:24data about your organization like how your company works, your org chart, maybe specific trade secrets that you don't want to be hosting on some third-party service. And you can do all of that internally with open weight models. And that's really one of the main benefits and one of the main misconceptions I want to clear up because this is a real benefit to US-based companies. If we can take these

12:43open weight models and modify them so that we can better do our business or whatever that may be, and not necessarily have to depend on a large AI company like OpenAI, Anthropic, or Google, really just do it ourselves, that also breeds innovation. We don't necessarily have to wait for these AI labs to make their own progress in terms of releasing newer and better models. We can take

13:04existing open models and then make them work really well for one specific purpose or one specific business case. And one of the last benefits of this do-it-yourself type of model is the fact that regular plain old people like you and me can actually run these models ourselves inside of our homes. And so, at that point, you don't even have to be connected to the internet. And if If have the right

13:25hardware, typically it's very powerful gaming machines, you can actually run these local models inside of your own home and never even touch the internet. And so, if we think about the different ways to access these models, as you can see, there are pros and cons to all of them, but it's not all bad. And that's really what I'm trying to get across here. If you're accessing open

13:44weight models from a Chinese provider, then yes, your data is traveling to China. And if you have trade secrets or intellectual property, you probably shouldn't be doing that. >> [music] >> But, if you have your own infrastructure or maybe a business agreement with Microsoft Foundry or something like that, then running these Chinese models actually doesn't introduce that same risk as sending your data over to China. Now, let's get into what open weight