Computer scientists are questioning whether Alphabet’s DeepMind will ever make A.I. more human-like

Published

4 years ago

June 18, 2021

admin

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.

Technology

Amazon faces FAA probe after delivery drone snaps internet cable in Texas

Published

4 hours ago

November 26, 2025

admin

Amazon faces FAA probe after delivery drone snaps internet cable in Texas

Amazon’s new MK30 Prime Air drone is displayed during Amazon’s “Delivering the Future” event at the company’s BFI1 Fulfillment Center, Robotics Research and Development Hub in Sumner, Washington on Oct. 18, 2023.

Jason Redmond | AFP | Getty Images

Amazon is facing a federal probe after one of its delivery drones downed an internet cable in central Texas last week.

The probe comes as Amazon vies to expand drone deliveries to more pockets of the U.S., more than a decade after it first conceived the aerial distribution program, and faces stiffer competition from Walmart, which has also begun drone deliveries.

The incident occurred on Nov. 18 around 12:45 p.m. Central in Waco, Texas. After dropping off a package, one of Amazon’s MK30 drones was ascending out of a customer’s yard when one of its six propellers got tangled in a nearby internet cable, according to a video of the incident viewed and verified by CNBC.

The video shows the Amazon drone shearing the wire line. The drone’s motor then appeared to shut off and the aircraft landed itself, with its propellers windmilling slightly on the way down, the video shows. The drone appeared to remain in tact beyond some damage to one of its propellers.

The Federal Aviation Administration is investigating the incident, a spokesperson confirmed. The National Transportation Safety Board said the agency is aware of the incident but has not opened a probe into the matter.

Amazon confirmed the incident to CNBC, saying that after clipping the internet cable, the drone performed a “safe contingent landing,” referring to the process that allows its drones to land safely in unexpected conditions.

“There were no injuries or widespread internet service outages. We’ve paid for the cable line’s repair for the customer and have apologized for the inconvenience this caused them,” an Amazon spokesperson told CNBC, noting that the drone had completed its package delivery.

Amazon delivery drone snaps internet cable in Texas

The incident comes after federal investigators last month opened a separate probe into a crash involving two of Amazon’s Prime Air drones in Arizona. The two aircrafts collided with a construction crane in Tolleson, a city west of Phoenix, prompting Amazon to temporarily halt drone deliveries in the area.

For over a decade, Amazon has been working to realize founder Jeff Bezos’ vision of drones whizzing toothpaste, books and other goods to customers’ doorsteps in 30 minutes or less. The company began drone deliveries in 2022 in College Station, Texas, and Lockeford, California.

But progress has been slowed by a mix of regulatory hurdles, missed deadlines and layoffs in 2023 that coincided with broader cost-cutting efforts by Amazon CEO Andy Jassy.

The company has previously said its goal is to deliver 500 million packages by drone per year by the end of the decade.

The hexacopter-shaped MK30, the latest generation of Amazon’s Prime Air drone, is meant to be quieter, smaller and lighter than previous versions.

Amazon says the drones are equipped with a sense-and-avoid system that enables them to “detect and stay away from obstacles in the air and on the ground.” The company recommends that customers maintain “about 10 feet of open space” on their property so drones can complete deliveries

The company began drone deliveries in Waco earlier this month for customers within a certain radius of its same-day delivery site who order eligible items weighing 5 pounds or less. The drone deliveries are supposed to drop packages off in under an hour.

Amazon has brought other locations online in recent months, including Kansas City, Missouri, Pontiac, Michigan, San Antonio, Texas, and Ruskin, Florida. Amazon has also announced plans to expand drone deliveries to Richardson, Texas.

Walmart began offering drone deliveries in 2021, and currently partners with Alphabet’s Wing and venture-backed startup Zipline to make drone deliveries in a number of states, including in Texas.

WATCH: Amazon unveils satellite terminal for enterprise customers — but Starlink still dominates

Technology

CNBC Daily Open: Nvidia’s crown looks increasingly uneasy on its head

Published

5 hours ago

November 26, 2025

admin

CNBC Daily Open: Nvidia's crown looks increasingly uneasy on its head

Jensen Huang, chief executive officer of Nvidia Corp., during the Taiwan Semiconductor Manufacturing Co. (TSMC) sports day event in Hsinchu, Taiwan, on Saturday, Nov. 8, 2025.

Lam Yik Fei | Bloomberg | Getty Images

Uneasy lies the head that wears the crown.

Shares of artificial intelligence czar Nvidia fell 2.6% on Tuesday as signs of unrest continued rippling through its kingdom.

Over the month, Nvidia has been contending with concerns over lofty valuations and an argument from the “The Big Short” investor Michael Burry that companies may be overestimating the lifespan of Nvidia’s chips. That accounting choice inflates profits, he alleged.

The pressure intensified last week in the form of a potential challenger to the crown. Google on Nov. 18 announced the release of its new AI model Gemini 3 — so far so good, given that Nvidia isn’t in the business of designing large language models — powered by its in-house AI chips — uh–oh.

And on Monday stateside, Meta, a potential kingmaker, appeared to signal that it is considering not just leasing Google’s custom AI chips, but also using them for its own data centers. It seemed like Nvidia felt the need to address some of those rumblings.

The chipmaker said on the social media platform X that its technology is more powerful and versatile than other types of AI chips, including the so-called ASIC chips, such as Google’s TPUs. Separately, Nvidia issued a private memo to Wall Street that disputed Burry’s allegations.

Power, whether in politics or semiconductors, requires a delicate balance.

Remaining silent may shroud those in power in a cloak of untouchability, projecting confidence in their authority — but also aloofness. Deigning to address unrest can soothe uncertainty, but also, paradoxically, signal insecurity.

For now, the crown is Nvidia’s to wear — and the weight of it is, too.

What you need to know today

And finally…

Lights on in skyscrapers and commercial buildings on the skyline of the City of London, UK, on Tuesday, Nov. 18, 2025. U.K. business chiefs urged Chancellor of the Exchequer Rachel Reeves to ease energy costs and avoid raising the tax burden on corporate Britain as she prepares this year’s budget.

Bloomberg | Bloomberg | Getty Images

The UK’s Autumn Budget is coming

The run-up to this year’s U.K. Autumn Budget has been different from the norm because so many different tax proposals have been floated, flagged, leaked and retracted in the weeks and months leading up to Wednesday’s statement.

It has also made it harder to gauge what we’re actually going to get when Finance Minister Rachel Reeves finally unveils her spending and taxation plans for the year ahead.

— Holly Ellyatt

Technology

Workday stock slips on light quarterly margin guidance

Published

8 hours ago

November 25, 2025

admin

Workday stock slips on light quarterly margin guidance

Workday CEO Carl Eschenbach, right, walks to the morning session during the Allen & Co. Media and Technology Conference in Sun Valley, Idaho, on July 11, 2025.

David Paul Morris | Bloomberg | Getty Images

Workday shares slid more than 5% in extended trading Tuesday after the finance and human resources software maker issued quarterly margin guidance that came in below Wall Street projections.

Here’s how the company did in comparison with LSEG consensus:

Earnings per share: $2.32 adjusted vs. $2.18 expected
Revenue: $2.43 billion vs. $2.42 billion expected

The company forecast a fourth-quarter adjusted operating margin of at least 28.5% and $2.355 billion in subscription revenue, according to a statement. The StreetAccount consensus was a 28.7% margin and $2.35 billion in subscription revenue.

Workday’s revenue grew about 13% year over year in the quarter, which ended on Oct. 31. Net income of $252 million, or 94 cents per share, was up from $193 million, or 72 cents per share, in the same quarter a year ago.

Subscription revenue in the third quarter totaled $2.24 billion, with an adjusted operating margin of 28.5%. Analysts polled by StreetAccount had anticipated $2.24 billion in subscription revenue and a 28.1% margin.

During the fiscal third quarter, Workday announced artificial intelligence agents for analyzing employee performance testing financial health, and the company revealed plans to buy AI and learning software startup Sana for $1.1 billion. Also, activist investor Elliott Management said it had built a Workday stake worth over $2 billion.

Workday has seen its stock decline this year as pundits discuss the risk of generative AI tools threatening the growth prospects for cloud software incumbents. Company shares have fallen 9% so far in 2025, while the Nasdaq Composite index has gained 19%.

WATCH: Workday CEO Carl Eschenbach: There’s a narrative that AI is eating into software, that is false