Will the most interesting AI in 2027 be a LLM?

1kṀ3802

2028

86%

chance

ALL

By most interesting I mean a mix of most talked about, most profitable, one with the most potential and most powerful and some other stuff.

In 2016 it would have probably been AlphaGo
in 2020 probably AlphaFold (maybe GPT 3 would have to look at what got more attention at the time.)
In 2022 probably Chat GPT (Chat GPT seems to have had a much bigger impact than DALL-E 2)

In 2023 so far probably GPT 4

Comment if you have any suggestions for how "most interesting" should be defined.

If I think it will make the Market better might resolve to 50% if there are two equally interesting ones and one is a LLM etc.

Get

1,000

to start trading!

Pessoas também estão operando

Largest context window for an LLM from a top three AI lab in 2026?

Which company has the best LLM at the end of 2026? (Artificial analysis leaderboard)

Will a new form of AI surpass LLM’s before 2030?

49% chance

Will the first AGI be an LLM that emulates Nobel-prize-worthy scientific research?

26% chance

There will be one LLM/AI that is at least 10x better than all others in 2027

17% chance

In Jan 2027, will the top 3 leading AI labs be offering fewer than 15 flagship LLMs between them?

65% chance

Will LLMs still mostly preserve high ELO rankings of AI safety people at EOY 2028?

44% chance

Will OpenAI announce AGI before 2028 conditional on it centrally being an LLM?

38% chance

There will be a company which offers LLM or AI-based arbitration by 2030

83% chance

Will there be a state-of-the-art LLM that is NOT based on next raw token prediction before 2029?

Ordenar por:

I know this would be for 2026, but would Jev be a contender?
OR would, idk, astra or mythos be that for this year?

@Qubits I'd say Astra and Mythos are significantly ahead of Jev, but Jev would probably be in the top 10.

@justifieduseofFallibilism What are you using for "interesting"?

Arguably to some extent Deepseek v4.1 Flash would also be interesting, iirc open router has it as some of the top tokens.

I appreciate your response.

Does this include textless LLMs?

vendeu Ṁ15 YES

@RemNi I'm not super sure what that would mean, but if it's trained on spoken words e.g. YouTube that will count.

@justifieduseofFallibilism

https://ai.meta.com/blog/textless-nlp-generating-expressive-speech-from-raw-audio/

Yes, it's growing like doodle jump jumping higher and higher.

I am using the loose heuristic that's the category of “other“ it’s always bigger than one intuitively believes.

But attempting to define boundaries around a top contender: what if the biggest technology ends up being autonomous agents that use LLM is as one of several components?

@MatthewRitter I guess I'll also resolve probabilistically. So if the top AI is 50 LLM and 50 other I'll resolve 50%. And if it seems 90% other but uses LLM for 10% I'll resolve to 10%. But I hope to be able to resolve to YES or NO.