Leading researcher who works at OpenAI predicts some extremely aggressive improvements in the next 2 to 3 years

https://twitter.com/RichardMCNgo/status/1640568775018975232

Tweet content:

I predict that by the end of 2025 neural nets will:

  • have human-level situational awareness (understand that they’re NNs, how their actions interface with the world, etc)
  • beat any human at writing down effective multi-step real-world plans
  • do better than most peer reviewers

autonomously design, code and distribute whole apps (but not the most complex ones)

  • beat any human on any computer task a typical white-collar worker can do in 10 minutes
  • write award-winning short stories and publishable 50k-word books
  • generate coherent 20-min films

The best humans will still be better (tho much slower) at:

  • writing novels
  • robustly pursuing a plan over multiple days
  • generating scientific breakthroughs, including novel theorems (tho NNs will have proved at least 1)
  • typical manual labor tasks (vs NNs controlling robots)
2 Likes

I have some (mostly baseless) predictions about the timeline for this sort of advancement and it is nowhere near this aggressive. Richard Ngo is a smart guy and one of the best in the world at what he does. It’s pretty wild to see someone who knows the field so well give such strong predictions in such a confident manner.

I don’t see why any of these things would take more than a year and a half to reach. We underestimate the effect of exponential improvement.

I doubt humans will be better at writing novels for long. How many writing workshop tools need to be incorporated into prompts and content generation apps before LLMs can use them well?

Once we figure out how to plug existing research/experimentation systems into generative AI, I think we’ll start seeing scientific breakthroughs start flooding through. Right now we are limited, like data science was, at the expense of the mechanics of setting up the data/prompts, but platforms for this are dropping every day (plug for my sister project at Microsoft - Semantic Kernel).

2 Likes

This is a really ambitious and exciting prediction! Neural networks have been making rapid progress over recent years, and I believe that these predictions could very well come true. If so, it would certainly be a game-changer in many industries.

It does, however, seem that some tasks require more than just technical knowledge - for example writing novels or achieving scientific breakthroughs. Certain tasks appear to be better suited for humans as they require more creative and complex thinking, which is something AI still lacks in comparison to humans. We will get there though. no doubt.

What is creativity?