Major Changes in AI Regulation and Chip Strategy

0:00Today on the podcast, we have big news out of the White House. two separate events. One is that Trump has made some big personnel replacement in the copyright office. And the second one is in regards to US chip export limitations from the Biden administration that have been recently repealed by the Trump administration. So a bunch of interesting things on the copyright IP trademark side, and also on the chip export side, we're diving into the implications and what is happening this week. Before we do, I wanted to mention that my very own startup, AI Box, is now launched and is in beta. So our first product, which is called the AI Box Playground, essentially allows you to test all of the different top AI models.

0:45There's over 20 of them on the platform for$20 a month. So you don't have to have subscriptions to OpenAI, Anthropic, Cloud, Google Gemini, all of the different platforms. You get them all on the same all-on-one platform. And the cool thing that I love is we have a bunch of image models that you might not be super familiar with, but ones like Ideogram, OpenAI, Black Forest Labs, Ideogram, I've been absolutely impressed with what it's able to generate. And then we also have a feature where you can toggle back and forth between different AI models to get them to both run the same prompt. So whether that's for a text document or an image, and you can also compare the results of an AI model side by side and see what you could get if you have different AI models generating the same thing.

1:30You could look across, you can look at them, you can compare them side by side or in a four panel, et cetera. So a lot of really cool features that we've packed in here. If you're interested, the link's in the description. It's AIbox.ai. I would love to have you try it out. Let me know if you have any good feature suggestions we're adding and iterating on this rapidly. So I've been working on this for a couple of years, excited to have it launched. All right, let's get into what is happening in the US government this week. So the first thing I want to talk about is the Trump administration has officially rescinded Biden's AI diffusion rules.

2:03So what exactly does that mean? So there's been a bunch of rumors going on for the last couple of weeks. This is the Department of Commerce, the DOC. And essentially what happened was there was, you know, it was called the artificial intelligence diffusion rule. um biden put it into place it was supposed to come into effect in a couple days and essentially it just had a bunch of u.s made ai chip export limits now we've been seeing just in the last few weeks a bunch of manufacturers and a bunch of ai companies going to the white house and saying hey don't put these you know export limits in effect nvidia i think in particular was like it's going to be bad for um our company obviously but also they're like if if you do that then china is just going to develop their own chips and they're going to ship it to countries.

2:46And so it wasn't just like China, but there's a bunch of countries on there that some people said were questionable. So exactly how this worked is Biden's rule had the world essentially like put into three different buckets. There was tier one, tier two, and tier three. So in tier one countries, you have like Japan, South Korea, those types of countries. Those ones had no export restrictions. So Japan, South Korea, they were totally cool. They could grab as many American chips as they wanted. Tier two countries had countries like Mexico and Portugal, which were kind of added in there. These aren't like our political adversaries, right?

3:22Like Portugal, it's in the EU. Mexico, it's on our border. And essentially, those ones, for the first time ever, were going to have a bunch of chip export limits. They could only get a certain amount of chips that we would export. And then there was tier three countries like China and Russia. And those were going to have super strict controls. So this isn't that the whole thing is going to be replaced, but apparently they're going to put in some new things in the future. One thing that the DOC did mention, though, they put some guidance out and they essentially told companies that if they're using Huawei's Ascend AI chips anywhere in the world, that violates U.S.

3:59export rules. So it's kind of interesting. They're like they're really trying to flex the muscle, not let you use certain companies, say that you have to. This is kind of the way it was it was going for a while. And essentially, they warned that there was a bunch of consequences of letting US AI chips be used to train AI models in China. And this is kind of what they're trying to get away from. You can see people like Sam Altman have really started to pivot the US against China and deep seek and say, he is a democratic AI model and those are authoritarian AI models and we need the democratic AI models to win.

4:31So kind of interesting. I'll give you the quote that the US Secretary of Commerce, Jeffrey Kessler, said. He said, the Trump administration will pursue a bold, inclusive strategy to American AI technology with trusted foreign countries around the world while keeping the technology out of the hands of adversaries. I think pretty much what is going to happen. OK, well, he also said, at the same time, we reject the Biden administration's attempt to impose its own ill-conceived and counterproductive AI policies on American people. OK, what I pretty much think is happening here is I think there's still going to be some sort of export rules.

5:01Personally, I mean, I know you don't really care about my opinion, but personally, Mexico and Portugal, I don't see much of a reason to put like restrictions if they want to buy it, let them buy Nvidia chips. The one thing that I think is probably going to happen here, though, is that we're going to start seeing more of a country by country basis. Trump's negotiating tariff policy and deals with a bunch of different countries. And I'm sure this kind of AI chips thing could be a lever on that where he's like, look, you make a certain good deal with you, we'll let you have unlimited US chips, blah, blah, blah, blah, blah.

5:32So I think that might be a factor in this. And then probably another factor is like, it's they said they're going to replace it. So evidently, some of the things Biden came up with might not have been a terrible idea about Portugal and Mexico, restricting them probably wasn't a great idea. So we'll see what happens there. All right. And I mean, by great idea, I just mean for the profits of US companies. So I guess more money to America if we don't have restrictions on Portugal and Mexico. So anyways, and a bunch of other companies have bucketed into tier three or whatever. So the other thing that is making a lot of news is that Trump has just fired the copyright office director.

6:08There's a bunch of reports around AI that had just came out from there. And evidently Trump wasn't a fan of that. So what happened was Shira Perlmutter, so she was ahead of the US copyright office. She was actually hired or put into place during Trump's last presidency in 2020. But it wasn't, I don't believe it was by him who appointed her. So of course, there are many people that are very upset about this. Someone said Donald Trump's termination of the register of copyright is a brazen, unprecedented power grab with no legal basis. And then they're like, it is surely no coincidence. He acted less than a day after she refused to rubber stamp Elon Musk's effort to mine troves of copyrighted works to train AI models.

6:58And while this may be true, I think Elon Musk isn't the only one probably pushing for this. We've seen in recent, in the recent months, we've seen Sam Altman, formerly very, or I guess not formerly, Sam Altman, very strong Democrat, campaigned heavily against Trump in multiple elections. He has been over to the White House to say, you know, essentially they wanted to repeal a lot of the copyright things. I mean, it makes sense. He's got a vested interest in trying to get as much data for as cheap as possible. So I think Elon Musk, Sam Altman, I think we saw Anthropic and we saw even NVIDIA over there.

7:35A lot of people are kind of lobbying in this direction and you can see where the financial incentives lie. So I don't think that's any shock. What's interesting here is I think just the fact that in the copyright office, they just released a big report. And so this is kind of the reason why many people are saying she got fired. And the report essentially said that it is, quote, not possible to prejudge the, essentially then they said, you know, the outcome of individual cases. They said there's a limitation on how much AI companies can count in their, quote unquote, fair use as a defense for when they train models on copyrighted content.

8:14And they said, you know, particularly if they got this in an illegal way. Now, for me, I'm not, like scandalized by that. I think that's probably actually a great rule, right? Like if you're going and, you know, buying one paid account of the New York Times, then using it to go scrape all the paid content of the New York Times, I think that's pretty shady. And I think you probably get caught and get in trouble for it. But it seems like a lot of these AI companies have done similar kind of sneaky strategies. The problem at this point is just like the models have all this data in there and it's hard for them to figure out where it all came from and how to get it all out once they've trained models on it.

8:49There's a lot of drama around open AI training, probably off of all of YouTube. They're very shady about it. Miriam Marati was asked by the New York Times about it a couple years ago, was very shifty in her response saying she didn't know. And then of course, she quit the company. So like, just kind of, I think, absolved herself of anything there for their sorrow video generation model. In any case, I think evidently these AI companies have scooped up literally everything they can get their hands on and trained the models off of trying to get any sort of competitive edge and make their models better.

9:21And so this is what's happening. So this is essentially what their document said. It says, but making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access goes beyond established fair use boundaries. Then the copyright office essentially goes on to suggest that the government is not going to intervene right now. It's a little premature. You know, they said it would be premature at this time. So on the one hand, like, yes, I think don't get it illegally in shady ways.

9:57But at the same time, literally, the person that I think person that got fired probably isn't the end of the world because she literally said that it would be premature to have any sort of government intervention at this time. So there was no plan to do anything. They did express support for quote unquote licensing markets. I'm a big fan of this. Adobe is a great example of a company that I think is doing this well. They have, you know, if you were on one of their stock photo websites that they purchased or that they own, and they use that to train their Shutterstock or the Shutterfly image generation model, then you get a royalty when people use your artwork to generate images.

10:35And to be honest, it's actually really tricky the way that this royalty program works. It's not like a, it's not a piece by piece basis. It's actually, it's actually essentially, they just take like how many total images you have in the data set, and then they just divide it and give you like a small royalty. So it's, it's not very a scientific exact art. But I'm also part of, you know, some music distributors like Lander, for example, that has an opt in where you can opt in to allow your music to be trained in AM models. I think this kind of opt in is a good option. And I'm evidently Lander will either license the data or they'll make their own music generation model, which is interesting.

11:11So I'm all for these kind of licensing markets. It seems like they're kind of supporting them as well. I don't think that's a bad thing. Of course, there's tons of lawsuits against OpenAI and other people that have been doing this. And Elon Musk, Jack Dorsey, the owner of Square, former founder of Twitter, Jack Dorsey recently said, delete all IP law. I think that's probably a little extreme. But in any case, I think that there's definitely a big kind of debate going on and people are worried that it's going to stifle AI innovation in America and that that's going to slow down our American companies.

11:50Is this good? Is this bad? I'm not going to really weigh in on this. I will say that China's DeepSeek doesn't care about copyright and all the AI models in China don't care about scraping everything they can in any shady ways and using it. And those will inevitably be the best models if they're allowed to do that. So I think that's going to probably be the argument from the other side saying like, delete all IP law or whatever. It's just like China has and China will beat us. But of course, stealing everyone's stuff doesn't feel great. So those are the two sides of it. It's interesting to see what will come up, what sort of laws remain in place.

12:23Thank you so much for tuning into the podcast today. If you enjoyed it, make sure to go over and check out AIbox.ai, my own AI startup. I'm super excited that we're launching our beta. $19 a month. You can replace a ton of subscriptions to 11 Labs for audio, tons of image generators, tons of text models that you might've been paying for. You get them all in one place and chat with them all in the same chat feed. So it's a really cool product I'm really proud of. Let me know what you think. The link's in the description. Thanks so much for tuning in and I will catch you next time.