TheChek News Flow

Gapping up:

MDB +23.4%, CRDO +17.9%, CANG +6.2%, DRVN +5.8%, SB +3.1%, FCF +2.1%, APYX +2%, OOMA +1.9%, WBD +1.9%, BCS +1.4%, BLTE +1.3%, NVTS +1.3%, MIR +1.3%, AGRO +1%, WAB +0.9%, LAD +0.9%, AMZN +0.8%, CMC +0.8%, ACIC +0.8%, EHC +0.7%, AMRX +0.6%, BA +0.6%, AZN +0.5%

Gapping down:

JANX -39.9%, BWLP -6.9%, VSTS -6.7%, IREN -4.5%, BKV -3.3%, HUN -3.3%, EXK -3.2%, CART -3%, XPO -3%, HL -2.3%, VVV -0.9%, DASH -0.8%

OpenAI Co-Founder Sutskever Joins the Skeptics

As Sri and I head to San Diego for the annual Neural Information Processing Systems conference this week (get in touch if you’ll also be there!), we’re excited to learn more about reinforcement learning, the model training technique du jour at all the major AI developers.

There’s rising skepticism among researchers, including OpenAI co-founder Ilya Sutskever, about the effectiveness of RL and whether it can advance AI to the level of artificial general intelligence, on par with human experts in scientific research, healthcare and other domains.

Sutskever, who left OpenAI last year to start his own AI lab, explained in a rare interview on the Dwarkesh podcast why AI models are struggling to handle real-world tasks that aren’t part of the evaluations that researchers use when they develop the models.

He said researchers use RL to help the models ace the evaluations, but that doesn’t improve the way the models generalize, or handle a wide variety of tasks. (We covered this topic last week in the context of OpenAI versus Google.)

His lukewarm opinion of RL puts him in a similar camp as another OpenAI co-founder, Andrej Karpathy.

Sutskever also questioned the definition of artificial general intelligence, the concept of AI that outperforms humans at most economically valuable work. Sutskever said AGI should instead refer to AI that can learn how to excel at tasks after it starts working on them, much like a teenager who can learn how to drive in the real world after just a few hours behind the wheel.

Sutskever said nobody knows how to develop this kind of machine intelligence yet. And that’s why we’re entering what he calls an “age of research,” meaning smaller-scale experiments of new methods, rather than the “age of scaling” of existing methods by putting more computing power behind them.

The comments suggest Sutskever disagrees with the current approach used by many leading AI developers. That approach includes relying on people to evaluate thousands of AI responses to specific queries or giving models simulated versions of popular apps so they can try hundreds of times to complete tasks successfully in the apps. Billions of dollars have been spent on those techniques this year!

Still, he acknowledged that even if today’s advanced AI developers stall out, their technology can still make “stupendous” revenue along the way, though he cautioned that profits would be harder to come by as they “work hard to differentiate [from] each other.”

Sutskever also had a lot of thoughts on existential risks from AI—unsurprising since it was likely one of his reasons for leaving OpenAI in the first place, given his current lab’s focus on safe superintelligence. One potential solution for ensuring AI acts in the best interest of humans is to essentially merge human bodies with AI, which he referred to as a more advanced version of Elon Musk’s Neuralink efforts, to make sure humans know everything an AI is thinking and so the AI would want to protect the human body on which it depends, Sutskever said.

This sounds somewhat in line with Ray Kurzweil’s Singularity vision in which humans and AI will merge by 2045!

Before then, making AI that learns on-the-fly will likely require a completely different way of developing models, Sutskever said. For instance, he believes models are missing something similar to the emotions that guide human judgment and decision-making. Models also need a “value function,” something that humans have that tell us whether any decision will lead us to a good or bad outcome, he said.

If you’re wondering how long it might take to get to this kind of AGI, don’t hold your breath. Sutskever estimated it could take anywhere from five to 20 years.

Belgian authorities detained three suspects after ‘strong suspicions’ of wrongdoing in award of contract

Belgian police raided the EU’s diplomatic arm in Brussels and the College of Europe in Bruges as part of a probe into suspected fraud related to funding of junior officials, the bloc’s prosecutors said.

The raids on Tuesday morning, which resulted in the detention of three suspects, were prompted by “strong suspicions” of fraud in a 2021-2022 decision by European External Action Service (EEAS) to award the European Union Diplomatic Academy to the College.

Italian former diplomat Federica Mogherini, who had previously led the EEAS, became rector of the college in September 2020 and director of the diplomatic academy in August 2022.

The searches by Belgian police took place at the EEAS in the heart of Brussels’ EU quarter, several buildings of the College of Europe and at the houses of suspects, the European Public Prosecutor’s Office (EPPO) said in a statement. The investigation was supported by the European Anti-Fraud Office, it added.

“Prior to the searches, the EPPO requested the lifting of the immunity of several suspects, which was granted,” the prosecutor’s office said.

Spokespersons of the EEAS and the College of Europe did not immediately respond to a request for comment. A spokesman for the Belgian police declined to comment on an ongoing case.

The EPPO said the investigation, which was first reported by EurActiv, focuses on whether the college or its representatives were given prior notice of the EEAS tender to establish the diplomatic academy before its official publication.

“There are strong suspicions that, during the tendering process for the programme, article 169 of the Financial Regulation related to fair competition was breached,” the EPPO said. “That confidential information related to the ongoing procurement was shared with one of the candidates participating in the tender.”

The EPPO said the facts under investigation “could constitute procurement fraud, corruption, conflict of interest and violation of professional secrecy. The investigation is ongoing to clarify the facts and assess whether any criminal offences have occurred.”

Company is expected to promote from within, with parks chief Josh D’Amaro and TV head Dana Walden considered leading contenders

Disney’s DIS 2.20%increase; green up pointing triangle past attempts to pick a successor to CEO Bob Iger have been less than magical. Now the company is trying to write a happier ending.

Iger, who replaced Michael Eisner in 2005 following a shareholder revolt, has postponed planned retirement dates five times and helped push out a number of top executives poised to succeed him. Veteran insider Bob Chapek became CEO in 2020, only to be replaced by Iger himself in a dramatic 2022 corporate coup.

Disney’s board of directors has said it would announce in early 2026 its pick to succeed Iger. The company is widely expected to promote from within, with parks chief Josh D’Amaro and television head Dana Walden considered the leading contenders by employees and people who work with Disney.

The two executives met with Disney’s board this summer in Orlando, Fla., to discuss their visions for the company should they become CEO, according to people with knowledge of the meeting.

Whoever is named as the next CEO is expected to work alongside Iger for a while to learn the ropes before officially taking over, according to a person familiar with the board’s thinking. Iger’s contract expires next December.

One of the board’s top priorities for the current succession process is to ensure there is as little drama surrounding it as possible, with no major exits or brain drain when the new CEO is picked, according to people familiar with the matter.

The company recently extended the contracts of its general counsel, chief financial officer and heads of communications and human resources so they end between 2027 and 2029, according to public disclosures. Disney has also had contract extension talks with other top executives who report to Iger, including the people who operate key business units, one of the people familiar with the matter said.

The board itself has been tight-lipped. “I think that the best the board can do is run a really thorough process so that in the end, when a decision gets made, people feel very good that we have done our homework,” board member Carolyn Everson said at a Wall Street Journal Leadership Institute event in November.

D’Amaro, who chairs Disney’s experiences unit that includes theme parks, consumer products, cruise ships and videogames, is widely considered the leading contender to be the next CEO. His strongest rival is Walden, a veteran television executive who is co-chair of Disney entertainment, which includes streaming.

Supporters of D’Amaro say he is analytical, comfortable talking to Hollywood talent who’ve collaborated with the parks and a charismatic public ambassador for the Disney brand.

Theme parks have become Disney’s most important business in recent years as traditional television has faltered and streaming isn’t nearly as profitable as TV networks once were. D’Amaro has overseen numerous price increases at the parks while remaining popular with fans.

D’Amaro’s vision for Disney’s future includes giving videogames a bigger role at the company and integrating gaming technology throughout its creative processes, said a person familiar with his thinking. The Experiences boss, who has been with Disney since 1998, championed a $1.5 billion investment in Epic Games last year and has overseen the relationship with the “Fortnite” maker.

Walden is less of a Disney native, having joined when it acquired the entertainment assets of 21st Century Fox in 2019. But her backers say she is the company’s most experienced and accomplished creative executive. Hit film and TV content, they say, is the engine that powers all of Disney’s businesses.

She shares oversight of streaming and helped manage that business’s improving profit margins. She also has experience managing some of the company’s most delicate and important commercial and creative relationships, from the recent carriage dispute with YouTube TV to September’s benching and reinstatement of late-night host Jimmy Kimmel.

Veteran film executive Alan Bergman, who runs streaming with Walden and oversees brands such as Pixar, Lucasfilm and Avatar, is considered unlikely to replace Iger by employees and company partners.

ESPN head Jimmy Pitaro, Disney’s other top operating executive, has expressed he is in his “dream job” and doesn’t expect to become CEO given his lack of experience in scripted entertainment and parks, according to people with knowledge of the matter. His contract has multiple years left on it, according to one of the people.

Many employees have aligned behind-the-scenes with D’Amaro or Walden, but no candidate wants to be seen as engaging in politicking, said people close to the company. Even so, every move and public appearance by senior executives is interpreted through the lens of how they may be positioning themselves for a post-Iger world.

Disney’s board has considered outside candidates including videogame company Electronic Arts chief Andrew Wilson, The Wall Street Journal previously reported. But people close to Disney believe an internal executive will probably be promoted to CEO.

The board hasn’t said whether it might consider appointing co-CEOs. Disney has never had multiple leaders before, though it has at times had a CEO working alongside a powerful deputy—another possibility in the post-Iger configuration.

Whoever takes over will face many challenges including managing a $60-billion expansion of theme parks, growing a collection of streaming services that face intense competition and re-energizing key movie brands such as Marvel. They will also have to contend with continued cord-cutting that has weighed on the legacy TV business at Disney and its rivals.

Disney board chair James Gorman is leading the CEO selection process. When the board picked Chapek in 2020, Iger was chairman.

Gorman was widely praised for managing a CEO transition at Morgan Stanley that, like Disney, featured internal candidates jockeying for the position. Executives who didn’t get the top job stayed with the financial institution, and Disney is hoping to do the same with its leadership team.

Disney executive contracts typically last three-to-five years, according to people who have worked at the company. Bergman, D’Amaro and Walden were appointed to their current positions in early 2023.

Contract extensions could include expanded responsibilities or bigger titles for top executives who don’t become CEO, as well as compensation increases, to motivate them to stay, according to people close to Disney.

The coalition backing the effort has appealed to Elon Musk for help funding the measure

A coalition that tried and failed to block OpenAI’s conversion earlier this year is back with a new tactic: a California ballot initiative aimed at reining in the startup’s power.

The planned initiative, dubbed the California Charitable Assets Protection Act, was filed Monday with California’s attorney general. It doesn’t mention OpenAI by name, but calls for the creation of an oversight board empowered to review and potentially reverse conversions to nonprofit organizations engaged in scientific and technological research that have happened in the state since January of 2024.

In late October, after a year of negotiating with the attorneys general of California and Delaware, OpenAI announced that it was converting its for-profit subsidiary to a public-benefit corporation.

The conversion was key to OpenAI’s fundraising efforts. The startup has seen explosive growth in recent years and has placed a host of expensive and high-profile bets on the future of AI, consumer devices and computing hardware.

OpenAI’s ascendance has drawn scrutiny from some AI safety groups, critics of Silicon Valley’s quest for superhuman intelligence and the startup’s business foes. The ballot initiative is backed by the Coalition for AI Nonprofit Integrity, or CANI, which earlier this year tried to block OpenAI’s conversion, saying it was a betrayal of the startup’s mission and would “unduly concentrate power in the hands of a few.”

“This is a baseless attempt by CANI to relitigate a decision that has already been made,” said an OpenAI spokeswoman.

OpenAI was co-founded by Sam Altman and Elon Musk in 2015 as a nonprofit organization with the goal “to advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.”

CANI is appealing directly to Musk for help funding the measure, which would appear on the ballot in November of next year if its proponents can gather the hundreds of thousands of signatures required. Musk, who left OpenAI in 2018, has tried to block OpenAI’s conversion in court.

The coalition said it has the initial funding to get the ballot measure in front of voters, “but to see it through to November 2026, we’re appealing to Elon Musk and others who understand what’s at stake.” Musk didn’t immediately respond to a request for comment.

To gather support, the coalition is launching a website called OpenTheft.com to raise funds, organize and spread its message. “We can’t do this alone. This is going to take resources to fight one of the most well-funded companies in the world,” the coalition said.

The paperwork for the ballot initiative was filed by Poornima Ramarao, the mother of an OpenAI whistleblower who died shortly after publicly accusing his former employer of breaking copyright law to train its models. The San Francisco medical examiner determined he had died by suicide.

OpenAI has said the company was “devastated” by Suchir Balaji’s death and called him a “valued member of our team.”

CANI tried earlier this year to push forward a bill in the California legislature that would have blocked OpenAI’s conversion. It sought to “prevent any individual, business or group from creating a startup venture capital nonprofit to exploit the charitable contributions of contributors and stakeholders, including knowledge, resources, and donations.”

News Corp, owner of The Wall Street Journal, has a content-licensing partnership with OpenAI.

A prominent European EV repair specialist is sounding the alarm on Tesla Model 3 and Model Y vehicles equipped with LG battery cells manufactured in China, claiming they are seeing “catastrophic” failure rates and significantly shorter lifespans compared to Panasonic packs.

For years, the narrative around Tesla’s move to Chinese battery suppliers has been generally positive, with the LFP (Lithium Iron Phosphate) packs from CATL proving to be extremely durable.

However, Tesla also sources Nickel Manganese Cobalt (NMC) cells from LG Energy Solution’s Nanjing facility for its Long Range and Performance models in Europe and parts of Asia.

Now, EV Clinic, a Croatia-based independent research and repair facility known for diving deep into battery diagnostics, has issued a severe warning regarding these specific LG NCM811 packs.

According to the firm, data from its repair center suggests a stark difference in quality between Tesla’s two main higher-energy-density packs: the US-made Panasonic NCA packs and the Chinese-made LG NCM packs.

“We are raising serious concerns about Tesla Model 3/Y LG NCM811 battery packs (LGES Nanjing), which exhibit very high failure rates and significantly shorter lifespans compared to Panasonic NCA packs (Made in USA).”

The shop claims that while Panasonic packs are generally repairable and can last up to 250,000 miles before cell failure, the LG equivalents are approaching end-of-life at around 150,000 miles.

More concerning is the nature of the failure. EV Clinic states that in over 90% of the cases they see with LG packs, cell-level repair is “impossible.”

The issue appears to be widespread degradation across the modules rather than a single bad cell bringing down the pack. They found that LG cells often show extremely high internal resistance.

“A failing Panasonic cell hits roughly 28 mΩ, which is the measurement for LG cells when brand new… Out of 46 cells, it’s common to find 15 cells over 100 mΩ ACIR, and the remaining 30 cells above 50 mΩ ACIR.”

The lab shared an example from a Tesla battery module:

Because the degradation is so uniform and severe, replacing a single faulty module is described as “operationally unsustainable,” as the remaining weakened cells are likely to fail in a cascade shortly after.

The situation has become so problematic for the shop that they announced they are introducing a “feasibility fee” just to check if these specific packs can be repaired, noting that they are “losing over €20,000 each month” attempting to fix packs that are effectively dead.

At this moment, during ongoing experimental testing with real customers experiencing LG failures, we are losing over €20,000 per month in operational time while investigating whether LG’s Chinese NCM811 systems can be sustainably repaired. At this stage, we can confidently say: the cells are, to put it mildly, catastrophic. Panasonic has mostly single-cell failures at 250,000km, and it is repairable, whereas LG has multiple-cell failures.

Their advice to owners with failed LG packs? Swap it for a used Panasonic pack or go to Tesla for a full replacement.

A Google Search exec said that one of the company’s biggest opportunities in AI lies in its ability to get to know the user better and personalize its responses.

The promise is AI that’s uniquely helpful because it knows you. But the risk is AI that feels more like surveillance than service.

In a recent episode of the Limitless podcast, Robby Stein, VP of Product for Google Search, explained that Google’s AI tends to field more queries that are advice-seeking or those where the user is looking for recommendations — and these types of questions are more likely to benefit from more subjective responses.

“We think there’s a huge opportunity for our AI to know you better and then be uniquely helpful because of that knowledge,” Stein said in the interview. “And one of the things we talked about at [Google’s developer conference] I/O was how the AI can get a better understanding of you through connected services like Gmail.”

Google has been integrating AI into its apps for some time, starting back when Gemini was still known as Bard. More recently, it began pulling personal data into another AI product, Gemini Deep Research. And Gemini is now infused into Google Workspace apps like Gmail, Calendar and Drive.

But as Google integrates more personal data into its AI — spanning your emails, documents, photos, location history, and browsing behavior — the line between a helpful assistant and an intrusive one becomes increasingly blurred. And unlike opt-in services, avoiding Google’s data collection may become harder as AI becomes central to its products.

Google’s pitch is that this deep personalization makes the AI far more useful. The idea is that Google’s AI technology could learn from the user’s interactions across Google’s various services, then use that understanding to make more personalized recommendations. For instance, if it learned that a user likes particular products or brands, the AI responses might favor those in its recommendations.

That, Stein said, would be “much more useful” than just showing users a more generic list of the best-selling products in a given category. “That is, I think, very much the vision — of building something that can be really knowledgeable for you, specifically.”

This idea isn’t all that different from how the “Others” in the hit Apple TV show “Pluribus” have gobbled up the world’s knowledge, including intimate details about individuals. When the system interacts with the show’s protagonist, Carol, it uses that data to personalize everything: cooking her favorite meals, adopting a familiar face to handle its communications with her, and otherwise anticipating her needs.

But Carol doesn’t find the personalized responses kind; she finds them invasive. She never consented to sharing her data with the hivemind, yet it knows her better than she’d like.

Similarly, it seems that avoiding Google’s data-gobbling ways will get increasingly difficult in the AI era, and if Google doesn’t get the balance right, the results could feel more creepy than useful.

(To be clear: Google does let you control the apps Gemini uses to make its AI more knowledgeable about you specifically — it’s under “Connected Apps” in Gemini’s settings.)

If you do share app data with Gemini, Google says it will save and use that data according to the Gemini privacy policy. And that policy reminds users that human reviewers may read some of their data and not to “enter confidential information that you wouldn’t want a reviewer to see or Google to use to improve its services.”

But as more data gets ingested into Google’s own hivemind, it’s easy to see how AI could make data privacy more of a gray area.

Google, however, believes it has a solution of sorts.

Stein says that Google will indicate when its AI responses are personalized.

“I think people want to intuitively understand when they’re being personalized — when information is made for them, versus when [it’s] something that everyone would see if they were to ask this question,” he said.

Stein noted, too, that Google could send a push notification to users when a product they had been considering after several days of online research becomes available or is on sale.

“There are all these ways that Google now, across modes, across kind of different aspects of your life, [is] being incredibly helpful to you…” he said. “And I think that’s more of how I think of the future of search than any one specific feature or single form factor.”