Tech Brew Ride Home - Wed. 06/05 – An Open Letter To OpenAI

Episode Date: June 5, 2024

As I said on that bonus episode with Alex Kantrowitz, the research side of OpenAI isn’t happy, and they’re starting to speak out. More details on AI at WWDC next week. More price hikes in digital media. Palmer Luckey can’t stop; won’t stop. And how CoreWeave became one of the biggest winners of the AI era. Links: OpenAI Insiders Warn of a ‘Reckless’ Race for Dominance (NYTimes) Apple Made Once-Unlikely Deal With Sam Altman to Catch Up in AI (Bloomberg) Twitch is raising US subscription prices for the first time (Engadget) Quest v66 Update "Significantly" Reduces Quest 3 Passthrough Distortion & Warping (UploadVR) How an upstart is using its Nvidia ties to challenge cloud computing giants (Financial Times) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech meme right home for Wednesday, June 5th, 2024. I'm Brian McCullough today. As I said on that bonus episode with Alex Cantorwitz, the research side of OpenA.I. isn't happy and they're starting to speak out. More details on AI at WWDC next week. More price hikes in digital media. Palmer Lucky Can't Stop Won't Stop. And how Corweave became one of the biggest winners of the AI era. Here's what you missed today in the world of tech. Just a quick note up top that a whole bunch of construction flared up this morning right. outside my window. I wasn't able to head into my office studio for various reasons, and I tried to

wait the construction out, but at this point, it's go or no go. So if you hear some faint stuff in the background, my apologies. A group of current and former OpenAI and Google DeepMind staff have signed an open letter warning of a culture of recklessness and secrecy at, quote, frontier AI companies. Quoting the New York Times. The members say Open AI, which started as a nonprofit profit research lab and burst into public view with the 2022 release of ChatGPT is putting a priority on profits and growth as it tries to build artificial general intelligence or AGI, the industry term for a computer program capable of doing anything a human can. They also claim that OpenAI has used hardball tactics to prevent workers from voicing their concerns about the technology,

including restrictive non-disparagement agreements that departing employees were asked to sign. Open AI is really excited about building AGI and they are recklessly racing to be the first there, said Daniel Kokatilo, a former researcher in OpenAI's governance division and one of the group's organizers. The group published an open letter on Tuesday, calling for leading AI companies including OpenAI to establish greater transparency and more protections for whistleblowers. Other members of the group include William Saunders, a research engineer who left OpenAI in February, and three other former OpenAI employees, Carol Wainwright, Jacob Hilton, and Daniel Ziegler. Several current Open AI employees endorsed the letter anonymously because they feared retaliation from the company, Mr. Cocatilo said one current and one former employee of Google Deep Mind, Google's Central AI Lab, also signed.

A spokeswoman for Open AI, Lindsay Held, said in a statement, we're proud of our track record providing the most capable and safest AI systems and believe in our scientific approach to addressing risk. We agree that rigorous debate is crucial given the significance of this technology and will continue to engage with governments, civil society, and other communities. around the world. A Google spokesperson declined to comment, When I signed up for Open AI, I did not sign up for this attitude of let's put things out into the world and see what happens and fix them afterwards. Mr. Saunders said. Some of the former employees have ties to effective altruism, a utilitarian-inspired movement that has become concerned in recent years with preventing existential threats from AI. Critics have accused the movement of promoting doomsday scenarios about the technology,

such as the notion that an out-of-control AI system could take over and wipe out humanity. Mr. Coquatillo 31 joined OpenAI in 2022 as a governance researcher and was asked to forecast AI progress. He was not, to put it mildly optimistic. In his previous job at an AI safety organization, he predicted that AGI might arrive in 2050. But after seeing how quickly AI was improving, he shortened his timelines. Now he believes there is a 50% chance that AGI will arrive by 2027 in just three years time. He also believes that the probability that advanced AI will destroy or catastrophically harm humanity, a grim statistic often shortened to P-Doom in AI circles, is 70%. At OpenAI, Mr. Kukotilo saw that even though the company had safety protocols in place, including a joint

effort with Microsoft known as the deployment safety board, which was supposed to review new models for major risks before they were publicly released, they rarely seemed to slow anything down. Eventually, Mr. Coquatillo said he became so worried that last year he told Sam Altman, CEO of OpenAI, that the company should, quote, pivot to safety and spend more time and resources guarding against AI's risks rather than charging ahead to improve its models. He said that Mr. Altman had claimed to agree with him, but that nothing much changed. In April, he quit. In an email to his team, he said he was leaving because he had, quote, lost confidence that Open AI will behave responsibly as its systems approach human-level intelligence. The world isn't

ready and we aren't ready, Mr. Coquatillo wrote, and I'm concerned we are rushing forward regardless and rationalizing our actions, end quote. On his way out, Mr. Coquatillo refused to sign OpenAI's standard paperwork for departing employees, which included a strict non-discharagement clause barring them from saying negative things about the company or else risk having their vested equity taken away. Many employees could lose out on millions of dollars if they refused to sign. Mr. Coquatillo's vested equity was worth roughly $1.7 million, he said, which amounted to the vast majority of his net worth, and he was prepared to forfeit all of it. A minor firestorm ensued last month after Vox reported news of these agreements.

In response, Open AI claimed that it had never clawed back vested equity from former employees and would not do so. Mr. Altman said he was, quote, genuinely embarrassed not to have known about the agreements, and the company said it would remove non-dismaragement clauses from its standard paperwork and release former employees from their agreements. In their open letter, Mr. Cocatilo and the former OpenAI employees call for an end to using non-disperagement and non-disclosure agreements at OpenAI and other AI companies. Broad confidentiality agreements block us from voicing our concerns, except to the very companies

that may be failing to address these issues, they write, end quote. Quoting Dar Obasanjo on threads, the problem with OpenAI since ChatchipD launched is that its behavior is not consistent with a research lab that's worried it might accidentally build Skynet or Ultron. It acted like a startup, that the employees, it hired under that premise are now loudly quitting is unsurprising, end quote. Ahead of the WWDC keynote coming up on Monday, Mark German has one more prognostication in terms of what to expect. His sources say Apple plans to offer AI features as an opt-in service and is working on LLM-powered robotic devices. Quote,

The Open AI Partnership is likely a short-to-medium-term relationship for Apple, said Dag Kitloss, a tech veteran who co-founded and ran the Siri business before it was acquired by Apple. But you can bet that they will be working hard building out their own competencies here, he said. Apple picked OpenAI as its inaugural AI partner for a few reasons, one of the people said. It got better business terms than Google was offering, and Apple believes that Open AI's technology is the best available on the market. Integrating Google AI into the iPhone also might have given the impression that Apple's biggest technology rival had beat it in a vital new area.

OpenAI, meanwhile, will get the gigantic exposure that comes with being deeply integrated into some of the world's best-selling smartphones and tablets. Still, Apple's involvement may bring new scrutiny to the safety and privacy concerns swirling around chat GPT. Depending on how deeply Apple plans to integrate the chat bot with its software, it also could mean that OpenAI has access to personal information, which could unnerve some users. But Apple is expected to offer its new AI features as an opt-in service, according to the people familiar with the matter. So where he customers could easily steer clear of them if they'd prefer. Apple also is looking beyond chatbots. It aims to use large language models, a key technology behind generative AI, to help power a pair

of robotic devices that it is secretly developing, the people said. That includes a tabletop robotic arm with a large iPad-like display. The company has also been working on a mobile robot that can follow users around and handle chores on their behalf. And it's looking to equip its AirPods with cameras and AI features. More immediately, there's the opportunity for Siri to finally live up to its potential, Kitloss said. That could bring some vindication to a company that hatched the dream of a smart personal assistant under co-founder Steve Jobs. There are no longer any technical constraints to realizing the original Siri Vision, Kitloss said, end quote. Another price rise, though maybe this one's a little more niche. Twitch says its Twitch tier one subscriptions in the U.S.

will increase from $499 to $5.99 per month on July 11th, the first price rise for U.S. subscribers. Quoting a gadget. Twitch is joining Spotify, Max, Peacock, Crunchyroll, EA, and other content services, and everyone's favorite corporate trend of raising subscription prices. The Amazon-owned company said on Tuesday that Twitch, Tier 1 subscriptions in the U.S. will increase from $499 to $599 in July 11th. This is the first time the monthly cost has gone up for American subscribers. As part of our efforts to help creators build and grow their communities worldwide. The following countries received subscription price adjustments as a part of local subscription pricing. The company wrote in a support article. In a separate X reply, the company clarified that streamers will still earn the same 50 to 70%

through Twitch's revenue sharing program, so they will earn more per subscription, likely the rationalization for the questionable it's for the creator's framing. However, streamers earning extra revenue depends on Twitch's subscription numbers staying the same or increasing. An unpopular price hike could lead to a loss of paying subscribers if enough people shirk the increase. Twitch had warned this day would come when the company raised subscription prices in Canada, Australia, Turkey, and the UK in February, Chief Monetization Officer Mike Minton added that a U.S. subscription increase would probably arrive sometime this year. And here we are. The company has had a rough 2024, and we aren't even at the halfway point.

Twitch laid off a reported 500 employees in January to, quote, cut costs and build a more sustainable business as CEO Dan Clancy admitted the company wasn't profitable. For good measure, it cut how much creators earn from Prime subscriptions, and then last month, it removed every member of its safety advisory council replacing them with Twitch ambassadors, which sounds an awful lot like community volunteers, end quote. One of the things that the meta quest has been leaning into lately, especially after the Apple Vision Pro launch, is being an AR device as much as it's a VR device. To that end, meta has rolled out an update to significantly reduce the Quest 3's pass-through distortion and warping, also an experimental wrist-mounted menu button option, quoting Upload VR.

The V-64 update in April improved Quest 3's pass-through dynamic range and exposure control to make screens readable and bright lights less blown out at the cost of some image vibrancy. With V-66, Quest 3's pass-through is being improved yet again, this time bringing, quote, significantly reduced pass-through distortions and improved hand alignment. Meta says it's rolling out the pass-through change gradually as a sub-update, so getting V-66 doesn't automatically mean you have the improvement yet. With V-66 audio from the web browser and other 2D apps, will now continue to play while you're in an immersive app.

That means you can, for example, listen to music while playing a VR game. Playback controls have been added to the Quest System menu called the menu control bar, the V-66 update introduces an experimental feature that replaces the hand-tracking menu gestures, opening your palm and pinching your thumb and index finger together, with wrist-mounted menu buttons. Some developers already use this location for hand-tracking UI, so if meta plans to one day make this the default, those developers may need to change their approach, end quote. Meanwhile, Palmer Lucky is feeling frisky, I guess, hot on the heels of that newfangled Game Boy, quoting Upload VR again.

Lucky says he will announce that he's working on a new headset at AWE 2024, which is taking place June 18th through 20th. We've reached out to Lucky for any specific details or teases about the headset he plans to announce at AWE. Given the unconventional nature of Lucky's existing hardware efforts from the Oculus Rift to handheld retro consoles and autonomous weapons systems, there's a lot of space for speculation about what direction this headset might take. We'll update this article if we hear back from Lucky. podcast appearance last year, Lucky praised Apple's strategy with Vision Pro as, quote, going after the exact right segment of the market that Apple should be going after, and said he thought

meta abandoned the high end of the market too quickly, not giving the technology enough time to progress through the usual adoption cycle. There's been this consistent push at Facebook to make VR something that everybody is interested in using, but I think it's come somewhat prematurely. I wrote a blog post years ago called Free Isn't Cheap Enough, and in it, I lay out an argument against the idea that the thing holding VR back is cost. There was a thing that I said right before I left. I pointed out in the marketing and targeting discussions that there was too much Starbucks and not enough Mountain Dew, end quote. He also described the Quest Pro as a directionally good step, poorly executed, pointing out that meta completely failed to capitalize on shipping the first

consumer standalone with eye tracking integrated, end quote. Finally today, who have been the biggest winners of the AI moment thus far? are, well, Open AI, obviously. Microsoft has become the most valuable company in the world by being seen as investors as being the furthest ahead, maybe with AI, and of course, InVIDIA. But there's one more really big winner. Have we ever talked about CoreWeave? A year ago, Nvidia invested $100 million in Corweave, valuing them at $2 billion. This month, Corweave raised $8.6 billion in debt in equity at a $19 billion valuation. So from $2 to $9,000. $19 billion in a year. That's quite a leap. Coding the FT. The jump in price highlights how New Jersey-based

Corwee founded in 2017 by former equity traders as a cryptocurrency miner has become another beneficiary of the booming demand for AI chips that propelled Nvidia to a $2.8 trillion valuation and boosted the stocks of the chipmaker's other partners such as Dell and Super Micro. Now the company is looking to put that capital to use by building facilities in the UK and Europe. we've rents out access to its coveted stash of invidia chips, including the in-demand H-100 and the forthcoming B-200 that run inside its data centers. Its facilities were designed to accommodate the particular requirements of high-performance computing, from high-speed networking between clusters of AI chips to liquid-cooled servers, said Michael Entruder, its chief executive.

The way that the cloud is going to be used for the next generation is very different than how the cloud was used in its inception 20 years ago, said Entrotter, calling the company the Tesla to Big Tech's Ford. And Trotter said it was, quote, incredibly hard, pitching to its earliest lenders, who had to become, quote, experts in a field that they didn't know anything about to the point where they could lend billions of dollars and take it to their investment committee to propose a new asset class, such as using Nvidia's graphics processing units as collateral. Coreweave had long since pivoted away from cryptocurrencies when the release of Microsoft-backed

OpenAI's ChatGPT in November 2022 unleashed a huge wave of demand for AI computing, seizing its opportunity, the company quickly scaled up its financing efforts. It raised more than $420 million in the first half of 2023 in equity, then another $2.3 billion in debt financing a few months later. Some existing shareholders sold $642 million worth of stock to Fidelity and others in December. Then last month, it struck two more deals to raise $7.5 billion in debt and $1.1 billion in equity. Corweave needed the funding to, quote, be able to scale large enough that we could support anybody who wanted to participate in the AI boom, said Introtter, no matter how many thousands of chips they needed. Like Amazon Web Services or Microsoft's Azure, CorWeave offers an alternative to companies

buying and maintaining their own servers, offering flexible access to computing power. But unlike AWS, which was founded in 2006 and can host an almost endless range of applications and data, Corweave's data centers serve a specific niche of clients with extremely high-performance computing demands from AI and drug researchers to media groups, end quote. There are strong rumors of a Corweave IPO happening as soon as early next year. Nothing more for you today. Talk to you tomorrow.