Trey Causey

Writing, advising, and seeking a next role in AI. Based in Seattle.

1 min

Source julian.ac

Failing to Understand the Exponential, Again

It may sound overly simplistic, but making predictions by extrapolating straight lines on graphs is likely to give you a better model of the future than most “experts” - even better than most actual domain experts!

More on the "AI progress has plateaued" / "AI progress is exploding" debate, this time from Julian Schrittwieser, an AI researcher at Anthropic.

Reasoning about exponential progress is extremely hard to do for humans. As Schrittwieser points out, we made the same cognitive errors at the beginning of Covid.

His conclusion: using the METR evals as one measure, we’re right on track and models continue to make exponential progress.

Share on XShare on LinkedInMarkdown for an LLM

Send your thoughts

Name and email are optional.

At least 10 characters.

Was this helpful or interesting?

All notes

ESC
Type to search...