Will LLMs be generally able to produce human-level text in 2026?

MANIFOLD

Ṁ100Ṁ109

Dec 31

55%

chance

ALL

such that IQ 140 person casually browsing the internet cannot tell what's human and what is not

Market context

Technical AI Timelines

OpenAI

AI Impacts

LLMs

Get

1,000

to start trading!

People are also trading

By 2027, will it be generally agreed upon that LLM produced text > human text for training LLMs?

61% chance

By 2029 end, will it be generally agreed upon that LLM produced text/code > human text/code for training LLMs?

77% chance

Will there be any simple text-based task that most humans can solve, but top LLMs can't? By the end of 2026

63% chance

Will LLMs' loss function achieve the level of entropy of human text by the end of 2030?

61% chance

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

44% chance

Will LLMs be used for academic peer review by 2030?

77% chance

Will LLM based systems have debugging ability comparable to a human by 2030?

71% chance

Will the most interesting AI in 2027 be a LLM?

84% chance

Will any widely used LLM be pre-trained with abstract synthetic data before 2030?

78% chance

In 2030, will LLMs be sending messages to coordinate with one another, whether we can decode them or not?

94% chance

Sort by:

@skibidist Such that a casual 140 IQ person cannot tell at what rate of success? And with how much scaffolding in place to try to coax the LLM output into seeming more human?

Title and criteria don’t match well. For example, if you were to hypothetically be able to identify LLM generated text by presence of emdash, that wouldn’t on its own make LLM generated text be of a lower quality.