Connect with us

Published

on

David Silver, leader of the reinforcement learning research group at DeepMind, being awarded an honorary “ninth dan” professional ranking for AlphaGo.
JUNG YEON-JE | AFP | Getty Images

Computer scientists are questioning whether DeepMind, the Alphabet-owned U.K. firm that’s widely regarded as one of the world’s premier AI labs, will ever be able to make machines with the kind of “general” intelligence seen in humans and animals.

In its quest for artificial general intelligence, which is sometimes called human-level AI, DeepMind is focusing a chunk of its efforts on an approach called “reinforcement learning.”

This involves programming an AI to take certain actions in order to maximize its chance of earning a reward in a certain situation. In other words, the algorithm “learns” to complete a task by seeking out these preprogrammed rewards. The technique has been successfully used to train AI models how to play (and excel at) games like Go and chess. But they remain relatively dumb, or “narrow.” DeepMind’s famous AlphaGo AI can’t draw a stickman or tell the difference between a cat and a rabbit, for example, while a seven-year-old can.

Despite this, DeepMind, which was acquired by Google in 2014 for around $600 million, believes that AI systems underpinned by reinforcement learning could theoretically grow and learn so much that they break the theoretical barrier to AGI without any new technological developments.

Researchers at the company, which has grown to around 1,000 people under Alphabet’s ownership, argued in a paper submitted to the peer-reviewed Artificial Intelligence journal last month that “Reward is enough” to reach general AI. The paper was first reported by VentureBeat last week.

In the paper, the researchers claim that if you keep “rewarding” an algorithm each time it does something you want it to, which is the essence of reinforcement learning, then it will eventually start to show signs of general intelligence.

“Reward is enough to drive behavior that exhibits abilities studied in natural and artificial intelligence, including knowledge, learning, perception, social intelligence, language, generalization and imitation,” the authors write.

“We suggest that agents that learn through trial and error experience to maximize reward could learn behavior that exhibits most if not all of these abilities, and therefore that powerful reinforcement learning agents could constitute a solution to artificial general intelligence.”

Not everyone is convinced, however.

Samim Winiger, an AI researcher in Berlin, told CNBC that DeepMind’s “reward is enough” view is a “somewhat fringe philosophical position, misleadingly presented as hard science.”

He said the path to general AI is complex and that the scientific community is aware that there are countless challenges and known unknowns that “rightfully instill a sense of humility” in most researchers in the field and prevent them from making “grandiose, totalitarian statements” such as “RL is the final answer, all you need is reward.”

DeepMind told CNBC that while reinforcement learning has been behind some of its most well-known research breakthroughs, the AI technique accounts for only a fraction of the overall research it carries out. The company said it thinks it’s important to understand things at a more fundamental level, which is why it pursues other areas such as “symbolic AI” and “population-based training.”

“In somewhat typical DeepMind fashion, they chose to make bold statements that grabs attention at all costs, over a more nuanced approach,” said Winiger. “This is more akin to politics than science.”

Stephen Merity, an independent AI researcher, told CNBC that there’s “a difference between theory and practice.” He also noted that “a stack of dynamite is likely enough to get one to the moon, but it’s not really practical.”

Ultimately, there’s no proof either way to say whether reinforcement learning will ever lead to AGI.

Rodolfo Rosini, a tech investor and entrepreneur with a focus on AI, told CNBC: “The truth is nobody knows and that DeepMind’s main product continues to be PR and not technical innovation or products.”

Entrepreneur William Tunstall-Pedoe, who sold his Siri-like app Evi to Amazon, told CNBC that even if the researchers are correct “that doesn’t mean we will get there soon, nor does it mean that there isn’t a better, faster way to get there.”

DeepMind’s “Reward is enough” paper was co-authored by DeepMind heavyweights Richard Sutton and David Silver, who met DeepMind CEO Demis Hassabis at the University of Cambridge in the 1990s.

“The key problem with the thesis put forth by ‘Reward is enough’ is not that it is wrong, but rather that it cannot be wrong, and thus fails to satisfy Karl Popper’s famous criterion that all scientific hypotheses be falsifiable,” said a senior AI researcher at a large U.S. tech firm, who wished to remain anonymous due to the sensitive nature of the discussion.

“Because Silver et al. are speaking in generalities, and the notion of reward is suitably underspecified, you can always either cherry pick cases where the hypothesis is satisfied, or the notion of reward can be shifted such that it is satisfied,” the source added.

“As such, the unfortunate verdict here is not that these prominent members of our research community have erred in any way, but rather that what is written is trivial. What is learned from this paper, in the end? In the absence of practical, actionable consequences from recognizing the unalienable truth of this hypothesis, was this paper enough?”

What is AGI?

While AGI is often referred to as the holy grail of the AI community, there’s no consensus on what AGI actually is. One definition is it’s the ability of an intelligent agent to understand or learn any intellectual task that a human being can.

But not everyone agrees with that and some question whether AGI will ever exist. Others are terrified about its potential impacts and whether AGI would build its own, even more powerful, forms of AI, or so-called superintelligences.

Ian Hogarth, an entrepreneur turned angel investor, told CNBC that he hopes reinforcement learning isn’t enough to reach AGI. “The more that existing techniques can scale up to reach AGI, the less time we have to prepare AI safety efforts and the lower the chance that things go well for our species,” he said.

Winiger argues that we’re no closer to AGI today than we were several decades ago. “The only thing that has fundamentally changed since the 1950/60s, is that science-fiction is now a valid tool for giant corporations to confuse and mislead the public, journalists and shareholders,” he said.

Fueled with hundreds of millions of dollars from Alphabet every year, DeepMind is competing with the likes of Facebook and OpenAI to hire the brightest people in the field as it looks to develop AGI. “This invention could help society find answers to some of the world’s most pressing and fundamental scientific challenges,” DeepMind writes on its website.

DeepMind COO Lila Ibrahim said on Monday that trying to “figure out how to operationalize the vision” has been the biggest challenge since she joined the company in April 2018.

Continue Reading

Technology

Nvidia says it will record $5.5 billion charge tied to H20 processors exported to China

Published

on

By

Nvidia says it will record .5 billion charge tied to H20 processors exported to China

Nvidia CEO Jensen Huang delivers the keynote address during the Nvidia GTC 2025 at SAP Center on March 18, 2025 in San Jose, California. 

Justin Sullivan | Getty Images

Nvidia said on Tuesday that it will take a quarterly charge of about $5.5 billion tied to exporting H20 graphics processing units to China and other destinations. The stock slid more than 6% in extended trading.

On April 9, the U.S. government told Nvidia it would require a license to export the chips to China and a handful of other countries, the company said in a filing.

The disclosure is the strongest sign so far that Nvidia’s historic growth could be slowed by increased export restrictions on its chips, which the U.S. government says can be used to create supercomputers for military uses. Nvidia reports fiscal first-quarter results on May 28.

During President Biden’s administration, the U.S. restricted AI chip exports in 2022 and then updated the rules the following year to prevent the sale of more advanced AI processors. The H20 is an AI chip for China that was designed to comply with U.S. export restrictions. It generated an estimated $12 billion to $15 billion in revenue in 2024.

Nvidia CEO Jensen Huang said on the company’s last quarterly earnings call in February that revenue from China had dropped to half of pre-export control levels. Huang warned that competition in China is growing, and for the second straight year, Nvidia listed Huawei as a competitor in its annual filing.

China is Nvidia’s fourth-largest region by sales, after the U.S., Singapore, and Taiwan, according to the company’s annual report. More than half of its sales went to U.S. companies in its fiscal year that ended in January.

Nvidia’s H20 chip is comparable to the H100 and H200 AI chips used in the U.S. and other countries, but it has slower interconnection speeds and bandwidth. It’s based on a previous generation of AI architecture called Hopper introduced in 2022. Nvidia is now focusing on selling its current generation of AI chips, called Blackwell.

DeepSeek, the Chinese company whose competitive AI model R1 unveiled earlier this year upended markets, used H20 chips in its research.

In addition to the existing Chinese export controls, Nvidia also faces new restrictions on what it can export starting next month, under “AI diffusion rules” first proposed by the Biden administration.

Nvidia has argued that further controls on its chips would stifle competition and potentially even erode U.S. competitiveness in technology. The company previously said it moved some of its operations, including testing and distribution, out of China after the 2022 export controls.

At the company’s annual conference last month, when asked about Chinese export controls, Huang said Nvidia works to comply with the law, but he also noted that about half of the world’s AI researchers are from China, and many of those work at U.S.-based AI labs. 

Nvidia said in Tuesday’s filing that the U.S. government told the company on Monday that the license requirement for H20 chips would be in effect “for the indefinite future.”

Nvidia shares have dropped 16% this year, largely due to President Trump’s announcement of widespread tariffs on top trading partners. While exemptions have been made on various electronics products, including smartphones, computers and semiconductors, Trump and some officials said over the weekend that the reprieve was temporary and part of plans to apply separate tariffs to the sector.

Shares of Advanced Micro Devices dropped more than 7% in after-hours trading on Tuesday following Nvidia’s disclosure. AI chipmaker Broadcom fell almost 4%.

WATCH: Nvidia says U.S. requires license to export H20 products to China

Nvidia says U.S. requires license to export H20 products to China

Continue Reading

Technology

Figma confidentially files for IPO more than a year after ditching Adobe deal

Published

on

By

Figma confidentially files for IPO more than a year after ditching Adobe deal

Dylan Field, co-founder and CEO of Figma Inc., after the morning sessions at the Allen & Co. Media and Technology Conference in Sun Valley, Idaho, on July 11, 2024.

Bloomberg | Bloomberg | Getty Images

Design software maker Figma said on Tuesday that it has submitted paperwork to the U.S. Securities and Exchange Commission for an initial public offering.

The confidential filing lands 16 months after the company scrapped a deal to be acquired by Adobe for $20 billion due to regulatory pressure in the U.K. The San Francisco startup had originally agreed to the deal 2022. Adobe paid Figma a $1 billion termination fee.

Figma’s software is popular among designers inside companies who need to collaborate on prototypes for websites and apps. The company was valued at $12.5 billion in a 2024 tender offer.

“There are two paths that venture-funded startups go down,” Dylan Field, Figma’s co-founder and CEO, said in an interview with The Verge last year. “You either get acquired or you go public. And we explored thoroughly the acquisition route.”

The announcement lands at a precarious moment for the tech IPO market, which has been largely dormant since late 2021. The Trump presidency was expected to revive new offerings due to promises of less burdensome regulations.

But after filing their prospectuses with the SEC, fintech company Klarna and online ticket marketplace StubHub delayed their IPOs earlier this month following the market turmoil caused by Trump’s announcements on widespread tariffs. Digital banking service Chime, which had filed confidentially with the SEC, also postponed its planned offering.

Turo, a car-sharing service, withdrew its IPO prospectus in February, three years after filing its initial prospectus.

Figma was founded in 2012 and is backed by investors including Andreessen Horowitz, Durable Capital, Greylock Partners, Index Ventures, Kleiner Perkins and Sequoia Capital. The company, which ranked 26th on CNBC’s Disruptor 50 list in 2024, had about $600 million in annual revenue as of early last year.

— CNBC’s Ari Levy contributed to this report.

Don’t miss these insights from CNBC PRO

Figma CEO on failed Adobe deal, startup landscape, big redesign with AI

Continue Reading

Technology

Meta CEO Mark Zuckerberg considered spinning off Instagram from Facebook in 2018: FTC trial

Published

on

By

Meta CEO Mark Zuckerberg considered spinning off Instagram from Facebook in 2018: FTC trial

Thilina Kaluthotage | Nurphoto | Getty Images

Meta CEO Mark Zuckerberg considered spinning out Instagram in 2018 on concerns about the rising threat of antitrust litigation against Facebook, according to an email presented Tuesday in a Washington, D.C. courtroom.

During Zuckerberg’s second day of testimony in Meta’s antitrust trial with the Federal Trade Commission, lawyers representing the FTC introduced an email from May 2018, in which Zuckerberg appeared to comment on the possibility of separating the photo-sharing app his company purchased in 2012 for $1 billion.

“And i’m beginning to wonder whether spinning Instagram out is the the only structure that will accomplish a number of important goals,” Zuckerberg wrote in the email. “As calls to break up the big tech companies grow, there is a non-trivial chance that we will be forced to spin out Instagram and perhaps WhatsApp in the next 5-10 years anyway. This is one more factor we should consider.”

Facebook bought Instagram in 2012, when the photo app had 13 employees and Zuckerberg was poised to take his company public in what, at the time, was the largest tech IPO on record. The purchase of Instagram and 2014 acquisition of WhatsApp for $19 billion are at the heart of the blockbuster antitrust trial that kicked off Monday and could last weeks.

The FTC alleges that Meta monopolizes the social networking market, and has argued that the company shouldn’t have been able to complete those acquisitions. The agency is seeking to cleave the apps from Meta as a possible remedy.

Meta disputes the FTC’s allegations and claims the regulator mischaracterizes the competitive landscape and fails to acknowledge a number of rivals like TikTok and Apple’s iMessage, and not merely other apps like Snapchat. Earlier in the trial, the FTC also presented an Oct. 2013 email in which Zuckerberg told other Facebook executives that Snap CEO Evan Spiegel rebuffed his $6 billion offer to buy Snapchat.

Zuckerberg also said in the 2018 email that the company’s “best estimates are that, had Instagram remained independent, it would likely be around the size of Twitter or Snapchat with 300-400 million MAP today, rather than closer to 1 billion.” MAP is short for monthly active people.

WATCH: Mark Zuckerberg takes witness stand on first day of antitrust trial.

Mark Zuckerberg takes witness stand on first day of antitrust trial

Continue Reading

Trending