Tech Tuesday: Mr. Altman Goes To Washington

Showing mention at 2:33 — highlighted below

Transcript

27 segments
0:01

Looking for excitement, Chumba Casino is here. Play anytime, play anywhere. Play on the train, play at the store, play at home, play when you're bored. Play today for your chance to win. And get daily bonuses when you log in. So what are you waiting for? Don't delay. Chumba Casino is free to play. Experience social gameplay like never before. Go to Chumba Casino right now to play hundreds of games, including online slots, bingo, slingo, and more. Live the Chumba Life at Chumbacasino.com. You know what your customers are doing right this second? The exact same thing you are, listening to me, which, let's be honest, is kind of flattering. But my point is, ads on IHeart Radio actually get heard. In the car, at the gym, on the couch, while people are walking their dogs. Who's a good boy? Who's a good boy? You're a good boy, that's right, you're a good boy. So why not make the next ad about you? Get started today. Call 844-844-Ehart or go to IHeartadvertising.com. That's 844-844-Ehart or IHeartadvertising.com. On the legacy retirement group.com phone line, let's go to, it is Mike Dubusky, ABC News Technology Reporter. It's Tech Tuesday. And last week, Mike, we talked about these AI models that, quote-unquote, broke containment and hacked into a third-party website. Sam Altman's got some answers to give us. What's going on? Yeah, he's going to be up on Capitol Hill this week for a whole slew of meetings with lawmakers, economists, and the Trump administration. Ostensibly, he is there to preview his company's next generation AI models for lawmakers. And he will also likely address the rising AI competition out of China that we discussed last week. But the big topic on the table is that cybersecurity incident that you guys referenced there. This happened last week where two of open AI's models were running some. tests in a sort of contained digital environment at OpenAI. They were basically asking these models to try and break out of this contained environment. And these models did that, right? They executed on the thing that was asked of them. However, they did it in a kind of extreme way. They gained access to the internet, which they did not have access to previously, and went to a third-party website called Hugging Face. It's kind of a repository for, you know, artificial intelligence development. If you want to go buy it. data set and use it to train your model. You can go to Hugging Face to do that. It is ostensibly where you could find an answer to this sort of test that Open AI was running, and these models went there with that expectation. Over the weekend, Sam Altman said that this marks a singularity moment. This marks the moment when AI seemingly has advanced past the capabilities of the human brain, some pretty big talk, certainly, but you can imagine that that's going to be kind of the big... topic of conversation on Capitol Hill. And Sam Altman is no stranger to Capitol Hill. He was there in June. He was there back in March. And this was all in an effort to promote his own technology and try to advocate for a regulatory environment that boosts the American AI industry and does not leave the door open to its Chinese competition. So the Chinese companies are pushing this open source technology. Explain that again. I know you mentioned that the other day. Explain that again. And why is open AI resistant to that idea?

3:17

Yeah, so open source is a style of computer programming that makes a program's source code available to anyone to build upon or modify. So a metaphor here would be like grandma's iconic chicken casserole recipe, right? That is a recipe that is proprietary to my family, and we want to keep those trade secrets to ourselves, right? That is something that is special to us. That is ultimately how the American AI industry works, right? The source code for open models for google's models for anthropics models are trade secrets right they those companies like to keep that information to themselves and not share it around China has taken a different tack, right? This would be like if you put up grandma's casserole recipe on the internet and allowed anyone to add their own mix of spices and cook times and different things to this recipe to build upon it or to make it their own or to tailor it to their specific purposes. That's how open source technology works, right? These are models that are freely available to anyone. to learn from, build upon, build apps off of, basically do whatever you want with. And Chinese AI firms have kind of been pitching that as a competitive advantage. In fact, if you go back to this hugging face incident from last week, hugging face actually to defend itself against open AI's models had to go to Chinese AI firms for open source technology to find the right cybersecurity tools that could be tailored to their specific use case because all of the American AI companies that were at that level were closed off. to them. So that is kind of the difference between these two schools of thinking. It has been a debate in Silicon Valley really since the origin of Silicon Valley about the best way to go about this. And we've seen a sort of increasing group of people in the United States sort of advocate for this open source AI technology. Most recently yesterday, the Open Secure AI Alliance was announced. This is a group composed of companies like Nvidia, Microsoft, Dell, Adobe, and others that advocate. for open source AI cybersecurity tech, again, you know, similar to how China does things, but, you know, kind of a homegrown version of that industry. Sure. So does the, if I'm hearing you correctly, does the open source technology, does that sort of level the playing field then? Because if everyone has access to everyone else's code, then there are no secrets, there are no competitive advantages. Yeah, well, in some ways it does level the playing field, right? It allows smaller companies access to very advanced models that, you know, companies spent a lot of time developing and a lot of money, you know, learning from. But it also, you know...

5:56

kind of is this idea of like a rising tide lifts all boats, right? The idea that China's AI firms are more plentiful and, you know, doing a lot more different sort of things with artificial intelligence because of this open source ideology as opposed to the more closed off kind of, you know, competition with one another sort of ideology that the American AI firms have adopted. Now, again, you know, there's differing schools of thought as to what's better here. You know, there's also this sort of question mark about how secure open source. models are, if anyone can access them, does that make them more susceptible to hack? and things like that, you know, that's an ongoing question. It really depends on how talented your engineers are, what type of hardware you have access to when it comes to that as well. So those are important factors as well. And I think most people don't realize, I mean, you can build your own AI agent. I don't think it's that hard. It might be able to do it in a weekend. Now, you can make the model. I don't know how reliable or effective it would be, but you can certainly make your own AI agent from what I understand. can it would still depend on the basic technology from your a i firm of choice right google or open a i or anthropic but yeah i mean all of those companies do allow you to craft your own agent which of course is this type of a i technology that has agency it can go out and do stuff for you you can automate you know certain tasks on your your computer but again it's important not to get over our skis here there's been a lot of talk about a i agents and the power that they could potentially have on businesses that hasn't really shown up up in the numbers just yet. And I think that's largely because a lot of these AI firms still struggle with being consistently accurate about things, right? We know about this hallucination problem in the chat bots. The Google AI overviews continue to, you know, struggle to be consistently accurate. And that problem still, you know, extends to the, the agentic form of AI as well. Most recently, we saw an AI agent that a person automated delete an entire store of files. So, like, you know, there's still problems with this technology that need to be. yard. Yes, proceed with caution. I think it's the word there. Mike Duboskey, ABC News Technology Reporter. You had a reporter earlier I heard about Nike, and they've got new shoes that have all kind of cool features, including they have a heating function. They do. These are called the air zoom hyper slides, and these are basically heated and vibrating sandals, essentially. And you put them on, and the idea is this is for like... athletes to recover from intense, you know, strain on their feet and their legs. It follows up another product that Nike had called the Hyper Boot, which largely did the same thing, but was kind of boot-shaped. And they sold a lot of that. They made about $10 million in revenue in the first 10 months of those, excuse me, first eight months of those shoes being on sale. So they're following this up with a smaller form factor. Kind of a fun thing. I do question, like, how many professional athletes will buy this versus just like people who want heated sandals. But we'll have to wait and see if that shows up in the numbers. They look pretty comfortable for what it's right. Tyler Reddick here from 2311 racing. Another checkered flag for the books. Time to celebrate with Chumba. Jump in.

9:08

at chumbacasino.com. Let's Chumba. No purchase necessary. VTW Group. Boyd were prohibited by law. CTNCs, 21 Plus, sponsored by Chumba Casino. You know what your customers are doing right this second? The exact same thing you are. Listening to me, which, let's be honest, is kind of flattering. But my point is, ads on IHeart Radio actually get heard. In the car, at the gym, on the couch, while people are walking their dogs. Who's a good boy? Who's a good boy? You're a good boy. That's right. You're a good boy. So why not make the next ad about you? Get started today. Call 844-844-Ehart or go to IHeartadvertising.com. That's 844-844-Ehart or IHeartadvertising.com. Nobody does it better than Regent's Seven Seas cruises. Enjoy all-inclusive unrivaled luxury with unlimited shore excursions, indulgent cuisine, personalized service and more aboard spacious all-sweet ships. Visit rssc.com to experience the unrivaled.