Have VERSES and Karl Friston achieved a "breakthrough in active inference" that beats GPT-4 on BIG-bench before 2025?

1kṀ1026

resolved Jan 5

Resolved

ALL

VERSES, a technology company focused on building AGI by applying insights from Karl Friston's "free energy principle" (and employing Friston as Chief Science Officer), recently published an open letter that claimed the company's "Active Inference agents may soon be able to match and surpass the performance of leading AI state-of-the-art approaches based on deep learning while using orders of magnitude less data and fewer computations." The letter called on OpenAI to fulfill the assist clause in its charter and help VERSES develop and deploy safe AGI.

AGI

AI Race

Get

1,000

to start trading!

🏅 Top traders

#	Name	Total profit
1		Ṁ24
2		Ṁ16
3		Ṁ5
4		Ṁ4
5		Ṁ4

People are also trading

Will xAI develop a more capable LLM than GPT-5 by 2026

51% chance

Will certain contemporary publicly available GPT models be generally accepted as conscious at inference time by 2100?

31% chance

Will OpenAI announce a model with a name containing the string "GPT-4b" in 2025

16% chance

Will we have an open-source model that is equivalent GPT-4 by end of 2025?

82% chance

Will LLMs such as GPT-4 be seen as at most just a part of the solution to AGI? (Gary Marcus GPT-4 prediction #7)

91% chance

Will it be possible to disentangle most of the features learned by a model comparable to GPT-4 this decade?

37% chance

Will mechanistic interpretability be essentially solved for GPT-4 before 2030?

Sort by:

Thanks for setting up this market!

Giving them a year to beat GPT-4 on a benchmark is awfully generous. Besides, if you really want to bet yes on this, I recommend their penny stock under VRSSF as a better option.

@TychonNotos agreed. I'd add an algorithmic efficiency criterion, where they need to use under some amount of compute

predictedNO

@Nikola I'm more interested in whether they can match GPT-4 with levels of capital they'll actually spend than whether it's particularly compute efficient.

@Ryannn I see. You are basically looking for a proxy on whether they have a new technology which could put them in competition with existing AI companies, while I am suggesting a proxy for whether they have a breakthrough which legitimately justifies calling for OpenAI's assistance rather than it being a marketing ploy.