Musk said during a live interview with Nicolai Tangen, CEO of Norges Bank, on X, formerly Twitter, that AI is the fastest-advancing technology he has ever seen, with barely a week without a new announcement.
“My guess is that we will probably have AI that is smarter than any human being by the end of next year,” Musk said. “And then AI, the total amount of sentient AI calculations, I think, will probably exceed that of all humans in five years.”
Gary Marcus, founder and CEO of Geographic Intelligence, a machine learning AI startup acquired by Uber in 2016, disagrees with Elon Musk’s AI predictions. And he offered $1 million to prove him wrong.
Moments after the bet was published, Damion Hankejh, investor and CEO of ingk.com, offered to increase the bet to $10 million.
Hankejh did not respond to a request for comment.
Marcus said Musk hasn’t responded so far and this isn’t the first time he hasn’t heard back from the billionaire about a bet.
Marcus had already bet $100,000 against Musk’s claims that AGI was imminent. The bet grew to $500,000.
For Marcus, it’s not just about winning a bet. He told Business Insider that he would like to have a public discussion with Musk about the real status of AI.
“For years, many in the tech industry have been making claims that strike me as scientifically implausible,” Marcus told BI. “And there have been a lot of broken promises, but virtually no accountability.”
As an example, Marcus cited technologies such as driverless cars, saying it was easy to demonstrate but much harder to create a reliable device.
Marcus said he thinks people need to understand what’s actually realistic — and large language models won’t be smarter than humans in the next five years, he said. In fact, Marcus thinks we may be decades away from machines being as trustworthy and reliable in their thinking as humans.
Marcus said that while large language models are the best technology in AI right now, they’re not really that good.
Marcus cited hallucinations, security data leaks and reliability issues among those concerns. He also said the technology was used for disinformation and used copyrighted data to disenfranchise artists.
“We should want the world to find better and more reliable ways to develop AI,” Marcus said.
Marcus said one way to do this is through some sort of licensing authority requiring demonstration of the benefits of AI.
Musk has expressed his concerns about the technology. He said there was a 10-20% chance that AI would destroy humanity.
He also sued OpenAi, alleging that it betrayed its mission for the benefit of humanity. Marcus said he publicly defended him during the trial.
“I was actually pretty supportive of Musk,” Marcus said. “I think drawing attention to these things is a good thing.”
Marcus told BI that he believes AI could help humanity, but he is much less optimistic that AI will be reliable in the near future.
businessinsider
