Ever since the release of ChatGPT, OpenAI has been hinting that superintelligence might not be too far away, but now its CEO has given a rough timeline that he expects to play out.
OpenAI CEO Sam Altman has said that “superintelligence” will be here in a few thousand days. Now a few thousand days is up to interpretation — few thousand could mean anywhere between 3,000 to 5,000 days, which could mean that Altman believes that humanity will achieve superintelligence in 8-13 years. He said that he was able to make the prediction because deep learning — the technology that underpins LLMs like ChatGPT — simply “worked”.
“After thousands of years of compounding scientific discovery and technological progress, we have figured out how to melt sand, add some impurities, arrange it with astonishing precision at extraordinarily tiny scale into computer chips, run energy through it, and end up with systems capable of creating increasingly capable artificial intelligence,” he wrote on his blog. “This may turn out to be the most consequential fact about all of history so far. It is possible that we will have superintelligence in a few thousand days (!); it may take longer, but I’m confident we’ll get there,” he added.
“How did we get to the doorstep of the next leap in prosperity? In three words: deep learning worked. In 15 words: deep learning worked, got predictably better with scale, and we dedicated increasing resources to it,” Altman said.
Altman seemed to be saying that it appeared that deep learning would scale with more compute, and keep producing better and better models. LLMs like ChatGPT have already demonstrated some extraordinary capabilities, but if these models are trained for longer and longer, and on more powerful computers, they could keep getting smarter. This would mean that they would become ‘superintelligent’, and lead to some incredible advances.
“Although it will happen incrementally, astounding triumphs – fixing the climate, establishing a space colony, and the discovery of all of physics – will eventually become commonplace. With nearly-limitless intelligence and abundant energy – the ability to generate great ideas, and the ability to make them happen – we can do quite a lot,” Altman said.
There had been some uncertainty whether deep learning could simply be scaled to produce superintelligence. Some experts, like Meta’s Yann Lecun believe that LLMs themselves can’t reason, and no matter how much compute we add, they’ll never be able to be superintelligent. On the other hand, Anthropic CEO Dario Amodei has been saying that thus far scaling has worked, which means that unless they hit a roadblock, they can simply keep scaling LLMs to make them more intelligent. And with Sam Altman saying that superintelligence could be here in 8 to 13 years, the world as we know it could be turned on its head sooner than we think.