...
By Hugh Cameron
Reporter
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
The CEO of Google and its parent company Alphabet has issued stark warnings about users “blindly” trusting artificial intelligence, as well as the potential impacts should the financial “bubble” forming around the technology eventually burst.
In an interview with the BBC, Sundar Pichai said that AI was “prone to errors,” and warned against the exclusive reliance on chatbots for research. People, he told the outlet, "have to learn to use these tools for what they're good at, and not blindly trust everything they say."
Pichai went on to warn of “irrationality” in the current AI investment boom, and said that “no company is going to be immune” should there be a market correction at some point.
Read More on News
Why It Matters
The comments come ahead of the long-awaited launch of the next‑generation version of Google's DeepMind AI, Gemini 3.0, and reflect growing concerns among even AI’s most enthusiastic investors and pioneers that caution and safeguards are necessary to prevent it from getting out of control.
...
What To Know
Speaking with the BBC at Google’s California headquarters, Pichai called the growth of investment in AI an “extraordinary moment.”
Spending on AI among Big Tech firms is estimated to have reached $400 billion annually, and some projections say this could grow to $2 trillion a year by the end of the decade. High levels of investment in AI-focused firms has been credited with many of the stock market’s recent gains, but this speculative spending without tangible payoffs has concentrated value in a handful of AI-bullish tech stocks while raising fears that a correction could have significant consequences for the wider economy.
Pichai’s comments regarding the AI bubble “bursting” echo those made by OpenAI CEO Sam Altman, who in August told reporters, as reported by CNBC, that “investors as a whole are overexcited about AI,” while comparing current conditions to the dotcom boom of the late 1990s.
“I would say that lots of money is being invested in AI for reasons that aren't necessarily purely financial, and this is leading to some inflated valuations,” said William Quinn, co-author of Boom and Bust: A Global History of Financial Bubbles.
“It's true that many companies would be affected by a crash to some extent,” he told Newsweek, adding that the macroeconomic impact is likely to resemble the dotcom crash which primarily impacted tech firms rather the broad-based fallout of the housing market crash of 2007 to 2009.
“The most important variable is the extent to which a bubble is fueled by bank debt, and this one largely isn't,” he added.
Regarding the accuracy of current AI models, Pichai noted a disconnect between how fast the technology was developing and the safeguards being implemented to mitigate harmful consequences. An October study led by the BBC and coordinated by the European Broadcasting Union assessed the accuracy of ChatGPT, Copilot, Gemini and Perplexity, finding that these AI assistants “routinely misrepresent news content,” and that 45 percent of generated answers contained at least one major issue.
“I think it is very important to not blindly trust GenAI because they are extremely prone to hallucinate,” Sandra Wachter, a professor of Technology and Regulation at Oxford University and senior researcher with the Oxford Internet Institute, told Newsweek. She noted that such hallucinations were often “very subtle” and that the most harmful could be slight errors made by LLMS that users were unable to identify.
“I worry less about the –‘the capital of Germany is Rome’- type of hallucinations but the ones that are so subtle that people are not able to spot them,” she said. “The types that are not outrageously wrong but wrong enough to cause damage.”
François Chollet, an AI researcher and former Senior Staff Engineer at Google, told Newsweek that Pichai’s remarks ahead of the launch of Gemini 3.0 could be taken as “a sign of maturity in the industry.”
“If you market an LLM as an oracle or a superintelligence, you are setting yourself up for failure and liability,” he said. “If you market it as a creative tool or a 'reasoning engine' that requires human verification, you are being honest about the tool’s nature. Sundar is trying to align public expectation with technical reality, and that's a very good thing."
Pichai told the BBC that Alphabet was being "bold and responsible at the same time" in responding to consumer concerns about AI’s potential risks, while also increasing investments in AI security and implementing new means for users to detect whether images are AI-generated.
However, Chollet told Newsweek that regardless of the intent to mitigate AI risks, safeguarding LLMS remains “an unsolved computer science problem.”
“Most current safeguards are essentially filters or ‘instructions’ given to the model to refuse certain queries,” he said*.* “This is usually pretty brittle.”
What People Are Saying
**Google and Alphabet CEO Sundar Pichai told the BBC: "**We take pride in the amount of work we put in to give us as accurate information as possible, but the current state-of-the-art AI technology is prone to some errors."
Entrepreneur and former presidential candidate Andrew Yang, in a recent interview with Newsweek, said: "You're in a situation right now where a lot of the value is forward-looking and somewhat speculative. It's really not based upon current revenues, for example. And there are some huge companies that are making massive bets in this space and being rewarded for it in terms of their stock value."
“It certainly wouldn't surprise me if there was a significant correction at some point,” he continued. “At the same time, I 100-percent believe that AI is going to transform many industries and that there will be huge winners and losers."
Sandra Wachter, a professor of Technology and Regulation at Oxford University, told Newsweek: “If people use LLMs to ask questions they don’t know the answer to, how would they be able to spot a mistake? And now think what can happen if we implement these systems in areas where truth and detail matter e.g. finance sector, investment decisions, stock market or medicine or journalism.”
Brent Mittelstadt, professor of Data Ethics and Policy at the Oxford Internet Institute, told Newsweek: "What worries me in this situation more than anything is that as companies try to fix this problem through a range of safeguards (e.g., fine-tuning, RAG, pre-defined responses for certain prompts), they’re gaining more power to decide what is true or a ‘correct’ or ‘appropriate’ response to a prompt.
"This is a significant centralization of power to decide what is and is not true or which topics are appropriate to discuss, and how they should be discussed, and I don’t think these companies have the democratic legitimacy or strong regulatory constraints needed to exercise this sort of power responsibly. This isn’t to say they shouldn’t be trying to fix errors/hallucinations, but rather that doing so should be a much more open/democratic process than is currently the case."
Software engineer and AI researcher François Chollet told Newsweek: “When a user uses ChatGPT or Gemini for research, they often trust the output because it ‘sounds’ expert. They do not realize that the model is effectively improvising based on compressed internet data. Until the general public treats LLM output as a draft requiring verification, we will continue to see significant errors in decision-making.”
What Happens Next
Google is expected to release Gemini 3.0 in the coming weeks. Speaking at the Dreamforce AI conference in October, Pichai said that the latest iteration of the AI would come out before the end of 2025, and bettors on Polymarket place the odds of Gemini 3.0 launching by the end of the week at 94 percent.
The prediction market previously put these odds at 69 percent, which Pichai responded to on X with two "thinking face" emojis.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines