Advancements in artificial intelligence will increase more quickly than most people expect over the next few years, and yet, the technology is fundamentally unpredictable, Anthropic CEO Dario Amodei says.
Amodei made the comments on Thursday during a one-on-one conversation with Slack CEO Lidiane Jones at Salesforce's annual Dreamforce conference in San Francisco.
"You don't know what your AI system is capable of until it's deployed to a million people, and that's not a good thing. That's a bad thing, right? We'd love to know what your AI system is capable of before you before you launch it. And we've worked very hard on coming up with the right metrics and the right evaluations to do a better job of seeing ahead than any other company can," Amodei said, but "as with the safety, no one's perfect at this. The technology is inherently unpredictable."
Amodei also thinks that the speed of AI development is going so quickly that most people underestimate what its capabilities will become over the next two to three years.
"There's a release cycle here," Amodei said. "What we're going to see is something quite radical … We haven't fully absorbed how early we are, and how much there is to go. I think it's going to be super exciting, and generative AI will be even more central to, you know, business and enterprise discussions than it is this year. I think next year it will get more exciting."
Headquartered in San Francisco, Anthropic was founded in 2021 by several former OpenAI researchers and employees including Amodei, his sister Daniela Amodei, Jack Clark, Tom Brown, Jared Kaplan and Sam McCandlish.
The startup also raised a $450 million megaround in May, one of the Bay Area's largest venture capital rounds in the second quarter. A couple of months earlier, Anthropic publicly launched Claude, its generative AI chatbot.
OpenAI publicly launched its own chatbot, known as ChatGPT, in November, beating Anthropic by several months, but Amodei is unconcerned.
"We didn't end up being first to market, but we made sure that we got it right," Amodei said. "Similar to the iPhone, we're not first to market, but we hope that in the long run, we will be the best."
California Governor Gavin Newsom also made an appearance at Dreamforce and voiced support for the responsible use of AI.
If AI can "humanize" us and support empathy and collaboration instead of exacerbating polarization, Newsom said, "then it's worth the price of admission."
On Sept. 6, Newsom signed an executive order directing the state to research, analyze and prepare for the ethical and responsible use of generative AI, Politico reported.
OpenAI CEO Sam Altman has also been sounding alarms over the potential dangers of advanced artificial intelligence systems and has been on a global tour this year speaking to users and world leaders about the technology and the need for regulations.
On Tuesday, Altman said he hasn't seen anything particularly "scary" emerge yet in OpenAI's research, but "we know it'll come. We won't be surprised when it does."
The following day, Altman was in Washington D.C. to meet with lawmakers along with several other tech executives for the A.I. Insight Forum, an event that was organized by New York Senator Chuck Schumer.
Altman was joined by Tesla CEO Elon Musk, Meta CEO Mark Zuckerberg, Google CEO Sundar Pichai, Microsoft CEO Satya Nadella, Nvidia CEO Jensen Huang and Bill Gates, the NYT reported.
There has also been a flurry of other meetings and hearings in D.C. this week about AI.
On Tuesday, a Senate Judiciary subcommittee held a hearing titled "Oversight of A.I.: Legislating on Artificial Intelligence," which included Microsoft President Brad Smith and Nividia's Senior Vice President of Research William Dally as witnesses. And on Thursday, the U.S. Senate held a Commerce, Science and Transportation committee hearing titled "The Need for Transparency in Artificial Intelligence."
The legislative interest comes several months after Altman first addressed Congress during a May hearing at the Senate Judiciary committee.
