According to Anthropic chief David Amodei, it is only a matter of time before artificial intelligence surpasses humans. That could be the case as early as two years from now.
The Wall Street Journal managed to snare Anthropic CEO David Amodei for an interview at the World Economic Forum in Davos, Switzerland. In the interview, Amodei looks ahead to the development of the company’s Claude models and artificial intelligence in general. Technological advances will only accelerate: Amodei predicts that within two to three years, artificial intelligence could surpass the human brain.
“We don’t know exactly when it will come, but I don’t think it will be much longer than 2027 before AI systems are better than humans in almost everything,” Amodei says. That, according to Amodei, will be accompanied by a complete rethinking of how people value work but also themselves. “If we create AI systems that are good enough to replace human labor, how do we as humans still find meaning? We will have to figure that out.”
1 million chips
So much for the existential crisis facing us all. In the present, Anthropic is tinkering with its AI models in full force. According to Amodei, Web integration and a speech mode are on the horizon in the near future, and Claude will get a better “memory. The company is raising the necessary funds and, with AWS and Google, has two tech giants that are only too happy to open their moneybags for the company.
The smarter Claude gets, the more powerful chips Anthropic will need. Amodei thinks at least a million chips will be needed to power the AI models as early as next year.
AGI is marketing
Amodei deliberately does not use the term “artificial general intelligence” (AGI). This term is used in science to describe AI that performs any intellectual task at the same level as the human brain, but is now mostly appropriated by OpenAI. Amodei believes OpenAI has eroded the term into a “marketing term,” he told CNBC in an interview.
The Anthropic CEO has a point there somewhere. OpenAI is trying to commercialize AGI even before it effectively exists. It would eventually want to give Microsoft and other commercial partners access to AGI systems at the expense of its own promise never to market (too) intelligent systems. AGI, in OpenAI’s vision, is given more of an economic interpretation than a scientific one.
read also