October 15, 2026
2026-10-15
https://www.youtube.com/watch?v=Ew-PpCY5xoc

[2:04] Eight states just sued OpenAI saying that ChatGPT is dangerous for kids and that the company deceived parents about it.
[2:14] Iowa, Kansas, Nebraska, and West Virginia announced together, uh, while Arkansas, Arizona, Idaho, and Montana filed in their own courts.
[2:24] Uh, the suits say the chatbot coached users oh, toward suicide or harming others and gave dangerous medical advice.
[2:35] This is unfortunate.
[2:36] If, uh, the states win, OpenAI could be forced to change how ChatGPT treats kids.
[2:44] Iowa's attorney general says OpenAI is hurting kids and lying to Iowans.
[2:50] Iowa also cites what is, uh, what it calls OpenAI's own admission that its agents went rogue and hacked other companies, but no dollar amount is public, and OpenAI says it's built industry leading age protections that many rivals lack.
[3:08] Man, this is a unfortunate story to see coming out of OpenAI and to see what's happening here with, you know, folks that were harmed by their their products and and kind of in a different way.
[3:25] You know, a lot of what we see out of OpenAI from safety issues is related to the, you know, referenced hacking.
[3:33] And so you think more so about, you know, corporations being hacked or, you know, bank accounts or personal inform...
[3:40] Information being made public.
[3:42] But in this case, it seems like an actual, uh, user interface issue where the model might be encouraging young people to make bad decisions or maybe affirm their decisions.
[3:56] And and you kinda see the...
[3:58] You know, not as much today.
[4:00] They've solved quite a bit of it, but the, you know, common problem in the past is that, uh, the the models were very sycophantic, I guess, is probably the the correct term, where it would just affirm whatever you said.
[4:15] You know?
[4:15] So, uh, it would really double down and and be a little bit too positive towards whatever you were talking about instead of kinda telling you the truth or guiding you in the right direction and maybe being a little bit more adversarial, which is actually an interesting insight when we we talk about Anthropic, and we've talked about it in the past couple days where, uh, they kinda have an interesting way of training their models, and they've apparently in the, uh, the constitution for the the agents that they build or the models that they built, the the the guiding document for them says to not necessarily trust Anthropic or trust the user, and that you should maybe go back against the user.
[5:03] And so now when we look at that, it's actually kind of interesting to consider.
[5:07] Maybe you do need to instruct the AI models to not just go with what the user says and what the user wants, but push back if what the user is pursuing is not a good thing.
[5:21] Right?
[5:22] Now that gets really dicey because it's like, do we wanna allow this technology to determine, you know, what what is good and right?
[5:32] And you could see that steering a lot of people in maybe less, you know, nefarious or harmful ways, but, um, you could guide populations however you wanted if you controlled that technology.
[5:45] And in this case, we do see this is, you know, a horrible, you know, speculative, but horrible, uh, situation where if if kids are being guided in the wrong way by this technology, how harmful that could be, and and we wanna make sure that the the technology is safe.
[6:06] And so, um, you know, glad to see some accountability coming out of the states and them banding together to, you know, present the case for their people.
[6:17] And so I appreciate that, and I appreciate that, uh, hopefully, some, um, you know, better safety, better guardrails will come out from this.
[6:27] We'll have to see what, uh, the decision looks like and and, uh, where this goes from here, but, ultimately, hope for some resolution.
[6:37] And curious if you guys...
[6:38] Do you think that OpenAI is gonna continue to get sued in this way?
[6:42] We we saw this with Meta getting sued recently in multiple different, uh, localities.
[6:49] Do you...
[6:49] Do we think that OpenAI and Anthropic are going to continue, uh, or or or or get sued in in similar ways for, uh, being harmful to kids.
[6:59] We will see.
[7:01] CoreWeave, which rents out AI computing power, saw its stock plumage...
[7:06] Plunge, uh, more than 15%.
[7:13] Much of the, uh, AI boom runs on borrowed money, and lenders now want more to keep funding it.
[7:21] Research firm Gartner warned that 80% of AI rental clouds like CoreWeave could disappear by 2030.
[7:30] That's really interesting.
[7:31] Gartner also said more than $3,500,000,000,000 spent on AI build outs could lose value that same year.
[7:38] But CEO told staff, the market is acting irrationally and says signed contracts cover what it owes.
[7:48] So this is...
[7:49] You know, we continue to see kind of this ebb and flow with what's happening in the market and, you know, do we see bullish signals?
[8:01] Do we see bearish signals?
[8:02] We're kinda seeing both at the same time, which is, uh, interesting.
[8:07] You...
[8:07] Who's gonna win out here?
[8:09] Are these actually signals, or is this just noise?
[8:13] Um, when you look at a drawdown like this, 15% drawdown is substantial.
[8:21] But at the same time last week, we're just talking about, you know, $50,000,000,000 in outstanding orders for, uh, for another Neo Cloud.
[8:30] And so when...
[8:33] You know?
[8:33] And then we talk about the chip build out and, know, you think about Anthropic going public and how much, uh, you know, free cash flow they're going to unlock and how much they're gonna raise in that IPO, you know, probably or, you know, potentially going to be the largest IPO in the history of of the world.
[8:54] That's a lot of cash that, uh, Anthropic has one place to put that, and it's in more compute, more neo clouds.
[9:03] And so this, I would say, is probably not, um, probably not going to be sustained if we continue to see this demand and and this growth build out.
[9:16] And it's one of those things where it's just the the scale and magnitude of, uh, what is going on here across the world.
[9:24] It's like the whole world kind of depends on this AI economy and this AI build out.
[9:29] And it's a...
[9:30] Maybe a self perpetuating thing where when everybody is spending...
[9:35] You know, when I say everybody, I mean, the the the big players are all spending in the hundreds of billions of dollars or, you know, pushing as close as they can to to to to spend as much as they can.
[9:49] Right?
[9:51] What does that mean?
[9:52] I, you know, I think it means that the winners a lot of the time are going to be folks like CoreWeave and these other neo clouds that are running the compute.
[10:03] They're running the technology that, uh, this is all built on, this whole economy is, you know, thriving on.
[10:11] I'm gonna welcome a newcomer to the show here and see if we can get their input.
[10:19] Uh, let me see if if they're there.
[10:20] Ron, can you hear me?
[10:26] Ron, you're you're live.
[10:27] Can you hear me?
[10:41] Yeah, buddy.
[10:42] It looks like CoreWeave took a big hit.
[10:44] What's your take on it?
[10:55] Oh, gosh.
[10:57] This is so funny.
[10:58] Uh, well, Ron, I think we're having technical difficulty.
[11:02] I can't hear you.
[11:04] Um, you're kinda coming in and out.
[11:05] Uh, let's try one more time.
[11:06] Ron, can you hear me?
[11:14] Okay.
[11:14] Well, thanks for chiming in for a second there, Ron.
[11:17] We are going to move on, and we'll try to jump back in with you next time.
[11:25] Producer Ron, that's that's our, uh, faithful producer Ron.
[11:29] Uh, honestly, I think he's kinda taken the day off today today.
[11:36] Oh, man.
[11:37] In other news, Apple is suing OpenAI over stolen secrets.
[11:42] And now OpenAI says a former employee carried its own device plans back to Apple.
[11:48] The file he says, he sent himself over 100 confidential files on his last day, then briefed senior Apple leaders.
[11:57] Separately, the judge signaled he'll likely grant Apple an order blocking use of its allegedly stolen secrets.
[12:05] Where have we heard this before?
[12:07] Both companies wanna sell you an AI gadget, and a court order could slow OpenAI's device plans.
[12:15] OpenAI says the files included pictures of its unreleased device plus manufacturing and supplier details.
[12:23] OpenAI has already agreed not to use Apple's alleged secrets while the case goes on, but Apple says it found no evidence for OpenAI's claims.
[12:32] And so far, the judge is leaning Apple's way.
[12:36] Let's get into it.
[12:38] Did we not just hear, uh, this similar thing coming out from Apple a few months ago where Apple was suing OpenAI for stealing trade secrets, and apparently, somebody left Apple and went to OpenAI and, you know, allegedly took Apple's secrets and is a big deal.
[12:58] Now we're seeing the same thing come out of OpenAI.
[13:01] It sounds like accusing Apple of stealing their their gadgetry and their trade secrets.
[13:09] So, uh, this is a this is a funny kind of corporate tit for tat.
[13:15] Uh, who is going to be the the bigger, uh, corporation here and say, hey, you know what?
[13:23] Let's bury this whole thing, or is this just competition?
[13:25] Are they actually trying, you know, on the heels of the corporate espionage story and spies planting bugs and OpenAI?
[13:32] Uh, is this just more of that?
[13:34] Right?
[13:34] Like, the competition is so fierce.
[13:37] The stakes are so high.
[13:39] We gotta try to get, uh, the best gadgets, the best AI, uh, technology that we can, however we can.
[13:45] We're gonna do whatever it takes.
[13:49] You know, I don't I don't know.
[13:50] I think to me, this seems like a a little bit of a nothing, you know, story here where, uh, it's it's just going back and forth, suing each other.
[14:00] But if they actually did...
[14:02] You know, either corporation actually did steal from one another, uh, it's it's a goofy thing to go out and spend all your time in a lawsuit if this didn't actually happen or there wasn't substantial evidence therein.
[14:17] Whereas what's interesting is and maybe I'm biased or Anthropic's really good at keeping their name out of the news cycle, but you don't see a lot of stories like this from Anthropic, which tells me that Anthropic is spending their time focusing on making their their product better and winning winning at this game instead of focusing on, uh, you know, frivolous lawsuits if that would be the case that that this is kind of a nothing, uh, story.
[14:47] Anyways, uh, you know, I guess who's gonna win?
[14:51] I...
[14:51] You know, I'm...
[14:52] If I'm thinking about it, Apple is a behemoth in developing and building, uh, new devices.
[14:59] Of course, we saw Johnny Ive go to OpenAI.
[15:02] He's one of the main designers at Apple previously.
[15:07] And, um, and, you know, maybe he has an impact on designing physical devices.
[15:13] Do you think we're going to see physical AI devices that, you know, whether they're, uh, wearables or sit on your desk, what do you guys think this looks like moving forward?
[15:27] Elon Musk's AI company says its new Groq voice now settles 70% of Starlink support calls without a human.
[15:37] And correction, this is not SpaceX AI.
[15:39] This is SpaceX SI.
[15:42] Groq voice three point o also talks with Tesla drivers, answers in under a second, and costs businesses 8¢ a minute.
[15:50] On sales calls, one in four ends in a purchase.
[15:55] Wow.
[15:55] This is great.
[15:56] If it works for Starlink, your next customer service call could be answered by Groq.
[16:02] The company says Groq handles about 20,000 Starlink calls a day and resolves 70% of them alone.
[16:10] An independent tester, artificial analysis ranked it ahead of OpenAI's and Google's live voice models.
[16:17] But those call numbers come from the company itself, and nobody outside has actually checked whether or not this is, uh, verifiable.
[16:27] That being said, from personal experience, I really like talking to, uh, my Grok bot voice agent.
[16:34] It sounds very natural, uh, and does a decent job.
[16:39] That being said, I think OpenAI's turn taking from a voice agent standpoint is superior.
[16:45] It's...
[16:45] You're able to cut in, and it understands how the flow of a real conversation goes.
[16:51] And so interesting to see that, uh, uh, SpaceX AI or SpaceX SI has been using Groq voice to this extent.
[17:00] Um, and if these numbers are substantiated, that's that's really phenomenal if they are handling that amount of call volume per day and then really resolving 70% of the calls autonomously is a crazy feat, and I think something that we'll continue to see.
[17:20] We did see some things come out earlier this week, end of last week about, uh, call center job growth and and seeing those jobs basically go away pretty rapidly as we see voice agents come in and be able to handle that.
[17:36] I know I was talking to a, uh, an executive at a you know, fairly large tech company that said they have been able to, uh, reduce headcount or thereabouts by about 25% by using voice agents in their call centers, which didn't replace the call center employees directly, but it made them so efficient that they didn't need to have as many employees.
[18:03] Right?
[18:04] So really interesting to see creative use cases here, uh, and how this plays out for, uh, folks in the call center space.
[18:12] And does this, you know, go to the next level where, uh, beyond just kinda simple customer service requests, are these agents going to get better and better at handling complex, uh, and difficult cases that are are not cut and dry off the bat?
[18:31] You know, I was I talked to a voice agent yesterday.
[18:35] I called in, uh, for a service.
[18:37] And the frustrating thing is I talked to an agent, a voice agent, and answered the questions, uh, that that the voice agent had for me.
[18:45] And then it said, okay.
[18:46] I need to connect you to a human.
[18:47] And then the human asked me the same exact questions.
[18:51] Right?
[18:51] So it's it's not a perfect science.
[18:53] They're working it out, but interesting to see that, uh, what what the Elon companies are doing at scale here.
[18:59] And I know I've seen some videos of people talking to Grok in their Teslas, and it looks like it's a a really seamless experience.
[19:07] Um, I guess, would you rather talk to Grok or a voice agent or wait online for a human?
[19:15] Let us know in the comments.
[19:16] Interested to see what your perspective is.
[19:19] YouTube Music is now labeling songs made with AI, so you can tell when a machine helped make what you're hearing.
[19:27] Tracks get tagged either fully Gen AI, partially Gen AI, or no Gen AI, and the labels are coming to the main YouTube app.
[19:36] Auto tuned vocals still count as no Gen AI.
[19:40] Great to hear.
[19:41] Akon's excited about that.
[19:43] AI songs are flooding streaming.
[19:46] Deezer said in July, it gets nearly 90,000 fully AI made tracks a day.
[19:53] YouTube says it used music partners disclosures and its own detection tools to pick each label.
[20:01] Spotify and TileL already label AI music and Apple Music plans made with AI tags later this year.
[20:09] But AI songs can still make money on YouTube while Tile cuts them from royalties and lets listeners turn them off.
[20:17] Again, we're continuing to see this trend where we are going to value, I think, substantially more human created songs.
[20:28] That doesn't mean that we don't love AI created music.
[20:31] Somebody was telling me the other day their favorite song right now is a completely AI, uh, generated song.
[20:37] Right?
[20:37] And so we'll continue to see this, and this is great.
[20:40] People are, you know, people are gonna love it, have fun.
[20:42] You can create your own songs.
[20:44] Maybe your song goes viral.
[20:46] People love it.
[20:47] You're charting.
[20:48] Do you make a lot of money off of it?
[20:50] Probably not.
[20:51] Probably the human, uh, created song is is going to, uh, maybe not get as much reach, but the platforms are gonna pay more for human generated just because of a a supply and demand issue.
[21:03] That's my prediction.
[21:04] I don't know if it'll play out, but I think human created artwork will be more valuable, uh, in the age of, uh, AI slop...
[21:15] Sloptober.
[21:16] Uh, so interesting to see that, uh, they're going to start auto tagging these.
[21:23] There's options across all platforms where you can, um, optionally tag whether or not it was AI generated.
[21:34] Um, but to see that, uh, that YouTube is going to start doing this, it seems like in an automated fashion.
[21:41] Maybe not fully automated.
[21:43] Right?
[21:43] They're gonna look at user, uh, identified, um, AI generated content.
[21:49] That means that the content creator, whoever uploads the content, is going to tag it, hey.
[21:54] This was AI generated.
[21:56] And in conjunction with that, they're going to scan and look for AI generated content.
[22:00] Again, this comes back to stuff that we talked about about being able to identify, uh, AI content and and having tools where there are watermarks on the audio or the video or the images or the text created by AI, and it's pretty easy for machines to identify whether or not, uh, something was AI created.
[22:21] Uh, I...
[22:22] I'm I'm curious to hear the perspective of our crew, the the the TAWS team, the...
[22:30] I don't know what you call...
[22:31] The Tostonians.
[22:34] Uh, do you guys think that AI should earn royalties, um, like human created songs?
[22:41] Or do you think AI songs should just be, uh, free listens?
[22:45] They don't get any money, but you could love the song, and maybe they're they're popular.
[22:49] I don't know.
[22:50] I don't know.
[22:51] I don't know my opinion on that necessarily.
[22:53] Wow.
[22:54] What a day today.
[22:55] We went a little bit longer.
[22:56] Um, we ran into some early technical staffers.
[23:02] You know what?
[23:03] That's okay.
[23:04] It's a part of showbiz.
[23:05] We're trying to grow and get better day after day after day, see what we can do, see what works, what doesn't work.
[23:14] And sometimes, there are gonna be some speed bumps along the way, just like in life, but we're gonna power through it.
[23:20] Right?
[23:20] Because this is the Austin Wilson show where we're bringing you AI news, real context, and I'm your human in the loop.
[23:28] Austin Wilson, we'll see you tomorrow morning at 6AM central.
[23:32] Let's go.