Will a paper solely authored by an AI research agent receive at least 100 citations by EOY 2025?
➕
Plus
42
Ṁ3567
2026
15%
chance

Tokyo-based research lab sakana.ai has recently (as of August 2024) released an AI research agent which they claim is able to automate the entire ML research lifecycle - proposing experiments, writing code, running that code and collecting results. (https://x.com/SakanaAILabs/status/1823178623513239992)

At least in theory, this means that we have entered the era where an AI could plausibly publish a paper all by itself, without any coauthors.

This market resolves YES if, by the end of 2025, a scientific paper has been published in a peer-reviewed scientific journal with an AI (whether sakana's or not) as its sole author and that paper has received at least 100 citations in other papers published in peer-reviewed journals.

The paper does not need to be AI-related. Citation counts will be checked on OpenAlex: https://openalex.org

Get
Ṁ1,000
and
S3.00
Sort by:
bought Ṁ250 YES

Yes

Are you counting on this being a publicity stunt from a company without academic ethics in a scam journal designed to extract publishing fees which will get a high citation rate due to controversy and scandal?

No. I just think that an AI will write an article that will get published then be cited lots.

You know peer review in non-scam journals requires talking to the editors, responding to critiques from editors and reviewers and having someone guarantee the accuracy and novelty of the results? Who is going to sign it and handle the communications without putting their name on it?

Probably someone hired by the lab (private or academic) that developed the AI.

Why would they not include themselves as authors? If they provide funding, prompting, editing and submission that requires authorship.

Because the lab will want to show off that its AI can do research autonomously. The lab's name will be on the paper.

So a publicity stunt without conforming to academic ethics. Are there any non-scam journals which you are aware of where this is possible?

Yeah there are lots of small, non-scam journals. Some that even don't particularly care about whatever arbitrary "academic ethics" that's popular atm

Which ones accept LLMs as authors and also don’t require a corresponding author?

  1. we are talking about AI in general, not just LLMs

  2. the author doesn't need to be identified as an AI at submission

You expect it to be fraudulently submitted under a pseudonym, and then be revealed to be an AI for publicity?

Yes, that's a possibility I was getting at in my most recent comment.

It's also possible that it will be non-fraudulently submitted to a small, non-scam journal. There are lots of small journal with lax rules in e.g. philosophy.

Or a major journal could team up with the AI lab and even have an entire issue published of AI-written articles.

Which journals?

IDK, never paid that much attention and I forgotten by now. But u come across lots of them. Head of philosophy department where i took courses founded one for example. Very small, but also not scams. Anyway, I doubt the article that resolves this market will be from a small journal, it will probably be either an AI-not-revealed-to-be-AI, or a partnership between a lab and a major journal.

Unsurprising.

i mean who cares ? u meant to look for truth on this site not to pwn ppl. u ask me questions i politely respond.

I was wondering if I missed something. Based on this, I have not. Thank you for your time.

no worries i appreciated being able to solidify my thoughts

bought Ṁ50 NO

And also - clarifying what you take as sources of citation data is important. E.g. Google Scholar is relatively easier to game/manipulate and provides noticeable higher total citation counts than say Scopus/WoS. Among open sources OpenAlex is pretty decent in my recent experience (and also gives substantially lower citation counts than Google Scholar)

Fair enough. I based the 100 citations number on a listing of relevant AI papers by citation count, so I will check what citation sources that listing was using in order to maintain coherence.

The original listing did not clarify where they obtained citation counts, but their citation counts track OpenAlex most closely among providers I checked, so I will resolve based on them.

bought Ṁ50 NO

It might be good to bear in mind that a) the produced papers are trash and b) most (if not all) non-predatory publishers have explicit policies against AI authorship of papers (e.g. Springer: "Large Language Models (LLMs), such as ChatGPT, do not currently satisfy our authorship criteria (imprint editorial policy link). Notably an attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs." https://www.springer.com/gp/editorial-policies/artificial-intelligence--ai-/25428500)

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules