Technology

➚ MOVE FAST: 3D. Google DeepMind has released a new 3D world-building model, which could represent a huge leap forward in training robotic systems.

➘ BREAK THINGS: O3. OpenAI’s new o3 model hallucinates more often than its older o1 model, according to internal tests. While that’s in part because it makes more claims, OpenAI can’t completely explain why this phenomenon is happening.

Netflix has no China business, but that’s not a bad thing, co-CEO Ted Sarandos said today at Semafor’s World Economy Summit.

“Fifteen years ago, everyone thought [a presence in China] was existential,” he said. The company previously tried to reach the country’s more than 1 billion consumers through a deal with Chinese firm iQiyi. Chinese censorship laws meant that all Netflix programming had to pass a censor board, “and in three years, not a single episode of a single Netflix show cleared the censorship board.”

Since then, Sarandos said he has watched other media companies “grinding out all of their time to get there, and ultimately ended up in the same place I did, which is nowhere.”

AI isn’t alive — at least not by the human standards by which we define “life.” But the extent to which AI systems are conscious — even potentially suffering — is something that’s less universally agreed upon, even by developers of those systems. Dwarkesh Patel, a researcher and podcast host, is one of the first prominent tech voices to lend credibility to the argument.

In a podcast published last week, Patel connected the case of British abolitionists fighting slavery in the 1700s to his own concern that digital entities may be suffering from the work humans are requiring of them. “There are going to be trillions of digital beings,” he said. “I want it to be the case that they’re not tortured and put into conditions in which they don’t want to work and whatever. I don’t want galaxies worth of suffering.”

What “suffering” looks like for digital entities isn’t clear, but many advocates agree it’s a version of emotional pain or distress incomparable to what humans experience. It may be evidenced by erratic responses to large inquiries, among other things. If it doesn’t exist today, it may exist in the future, as developers approach AGI.

Every major model developer has vehemently denied any consciousness claim for their systems. But Patel’s comment suggests that the question of AI sentience is further entering the mainstream. Be ready to defend your position.

Catch up with the changing world around you — check out The Semafor View, our annual guide helping business leaders make sense of this moment. This year’s edition brings together bold provocations from world-class thinkers on the forces reshaping business, policy, and the global economy. From AI disruption to geopolitical realignment, The Semafor View is a must-read for leaders navigating today’s fractured world.

Elon Musk isn’t the only one fighting to keep OpenAI a nonprofit. More than 30 individuals, including Nobel laureates, former OpenAI employees, and law experts, today submitted a 25-page letter to the attorneys general of California and Delaware, asking them to block OpenAI’s proposed restructuring to a for-profit entity. While Musk’s argument revolves around a breach of contract claim, the central assertion remains — that OpenAI abandoning its nonprofit structure violates its original charitable mission.

“OpenAI is trying to build AGI, but building AGI is not its mission,” the letter reads. Its purpose, as stated in official documentation, is “to ensure that artificial general intelligence benefits all of humanity,” and not for the “private‬ gain of any person.”

It is legally complicated to change the mission of the company. “If there are inefficiencies, you can fix them,” Page Hedley, a policy and ethics advisor for OpenAI from 2017 to 2018 who signed the letter, told Semafor. “But if the problem is [that] investors don’t like that the mission comes first, that’s too bad. That’s baked into who you are.”

As the focus for AI development in the US shifts from safety to speed, the pressure to scale is intensifying. OpenAI says it can maintain its mission to benefit humanity as a for-profit entity, while its nonprofit arm will play a philanthropic role. “Our Board has been very clear: our nonprofit will be strengthened and any changes to our existing structure would be in service of ensuring the broader public can benefit from AI,” an OpenAI spokesperson told Semafor.

While OpenAI is one of the leading model developers, that status has also hampered its ability to meet some market demands. A portion of the investment in its latest funding round, which valued the company at $300 billion, came with the caveat that it transition to a for-profit company by the end of the year.

OpenAI is the only major AI firm not playing for profits. Simultaneously, it is, in many ways, already functioning like a for-profit entity through its capped-profit segment and massive commercial partnerships. That makes the status change murky. “If someone’s behavior is inconsistent with the law, you don’t change the law to match their behavior,” Hedley said. “You change their behavior to match the law.”

Change of tune. Social media has both positive and negative effects on society, but lately, young people are seeing more of the negative. Nearly half of US teens surveyed by Pew Research Center believe social platforms negatively impact people their age, up from 32% in 2022. A meager 11% of respondents see social media as mostly positive, less than half the amount who believed so a few years ago.

So what changed? In that time, TikTok has fought to remain in the states, and Elon Musk bought X, contributing to a rise of hate speech on the platform. AI-generated content (or “slop”) has found a home online, and platforms have leaned into selling the products that appear in users’ feeds.

Social media companies like Meta, YouTube, and TikTok have also rolled out a slew of parental controls and safety features for their younger users, but the survey responses indicate the updates haven’t improved teens’ experiences in any significant way.

It’s also noteworthy that while half of respondents believe social media negatively impacts their peers, only 14% said it negatively impacts them personally. Psychologists call this the third-person effect, where it’s easier to recognize consequences in others.

Social media companies are continuing to update their experiences, with the latest push being to integrate chatbots. xAI’s Grok is an active user on X, while Meta experimented with AI influencers before deleting them for customized chatbots. The next wave of social media appears to be AI-driven, but it’s unclear how much that will resonate with users.

How much Tesla’s profits fell in the first quarter of 2025 against the same quarter last year on anti-Musk backlash and weak car sales. The company stated that Trump’s tariffs and “changing political sentiment” could negatively impact future demand.

The European Union fined Apple and Meta a combined $800 million for breaking the bloc’s digital rules.

Apple faces a €500 million ($570 million) of those total fines for directing consumers away from competitors to its app store, while Meta was hit with a €200 million ($230 million) penalty for requiring EU users to pay to access ad-free versions of its products.

These are the first fines issued under 2024’s Digital Markets Act, which aims to curb the dominance of major digital platforms — and charge up to 10% of their global turnover if they don’t comply.

The latest fines are “relatively softball” compared to previous penalties Big Tech companies had faced from the EU, like a €4.3 billion hit on Google in 2018, Reuters noted — and risk giving the impression that EU Commission President Ursula von der Leyen has caved to pressure from the Trump administration. The lower fines could also be seen as a bargaining chip for the administration’s recently imposed tariffs.

“The nuance here is important, though,” Reuters editor George Hay notes, adding that while the timing of the fines may be strategic, the DMA is different from past harsher antitrust regulation, and it’s designed to encourage Big Tech companies to comply without immediately resorting to massive fines.

Specialist investors aren’t spooked by a recent plunge in oil prices and oil-company shares, and think the US is still headed for more drilling as President Donald Trump chases an “energy dominance” agenda, Semafor’s Tim McDonnell reports.

By pushing the oil price close to or below the breakeven point for many drilling companies, Trump is testing the patience and fortitude of industry leaders and their shareholders.

That pain will probably be short-lived: Recession aside, oil demand isn’t drying up anytime soon. And in the boom-bust world of oil economics, periods of underinvestment in drilling brought about by low prices are always followed by supply shortages that drive the price back up, Tim notes.