The last couple of years have seen the world be constantly amazed by the new-found capabilities of AI, but that growth might be slowing down.
Venture Capitalist and investor Marc Andreessen has said that AI models seem to be hitting a ceiling of capabilities. “There is this very interesting asymptotic kind of thing that’s happening right now,” he said on a podcast. “Two years ago, there was one LLM that was way out ahead of everybody else,” he said referring to ChatGPT. “And sitting here today, there’s, like, six, that are on par with that. And interestingly, at least for right now, they’re all sort of asymptoting at sort of the same point,” he added.
“They’re kind of hitting the same ceiling on capabilities. Now there’s lots of smart people in the industry working to break through those ceilings, but sitting here today, if you just looked at the data, if you just looked at the charts of performance over time, what you would say is there’s at least a local topping out of capabilities that’s happening,” Andreessen continued.
Andreessen’s thoughts were echoed by his partner in his VC firm Ben Horowitz. “If you look at like the improvement from GPT 2.0 to GPT-3 to 3.5, and then compare that from like GPT-3.5 to GPT-4, we’ve really slowed down in terms of the amount of improvement. And the thing to note on that is the GPU increase was comparable. So we’re increasing GPUs at the same like rate, but we’re not getting the intelligence improvements at all,” he said.
It’s an interesting observation. GPT-3 had burst on to the scene nearly two years ago, and stunned the world with its capabilities. Other versions of ChatGPT have brought with them more enhanced capabilities, but the pace of growth does appear to be slowing. And while some of the world’s smartest people are continuing to innovate — and have billions of dollars of fresh infrastructure to help them out — it might be while before we can get another step change in the capabilities of AI systems.