This is a market to track when the technological singularity at the quantitive level implied here will happen with LLMs.
A look into my thoughts behind this (Work in Progress): The technological singularity for me, and this is only one aspect of it, happens once no human being can keep up with the state of growth of how much technology is being released. If by 2030 for me.over 100k LLMs (the amount of LLMs individually released, fine-tunes and other such things count if they truly provide their own weight to the table, as if it were their own LLM) are released a day, that to me means we hit the Singularity and so i'm curious whether people think this is possible and when. Will post-sub-markets on when it gets more intense. Examples being "Will 100k LLMs be released every hour by 2040" or "Will 1M LLMs be released every day by 2040" implicative growth cycles.
Within reason, I'm well aware the term and evolution of "LLMs" will change drastically but i believe through in-depth conversation on this topic, it will be understood what we mean. Will engage equally and cautiously with every contribution before making my decision. Comment if you want me to do this with other technologies and other Singularity-related Timelines.
1,000
1.00@mariopasquato Yes they do because they can be turned into their own companies and products where each GPT’s outputs used is a possible LLM fine tuned independent of the main model it derived from. If LLMs can be treated as their own regardless of how they are interconnected, as long as they can in some general fashion be treated independently as their own version, I would count it.
@FlorisvanDoorn If you bet YES and there’s some evidence to back up that it will happen after 2030, then it will resolve YES. If there isn’t even evidence of that that can be given until end of this market, it resolves to NO, until the next year interval with the same poll. I want to be able to track this trend across specific intervals which is why I’m doing it like this, treating each market as an informational nexus. Will be providing evidence for both YES and NO across the board that I find so it can be deliberated.
@CDBiddulph I'm curious, why do you think no? I bet YES because i think it can be shown by end of 2024 that the amount of LLMs released and how many LLMs people say they are working on publicly, if looked at further, and reached out to, that we would see, in the US, something like at least 10,000 LLMS released on some open-source forum per day and the issue is more of a lack of good discoverability and not growth. I plan on attempting to conduct said study and post it here and another one questioning the factors taken into account as possible study for NO. Criteria will also be open and collaborative. If by end of 2024 it resolves either way then there will be an extensive look at the trends coming on people can learn from and further their own markets into more nuanced forms of precision, to which then ill repost this same poll again until 2030, with each previous market referenced and its information gathered.
@ThePhilosopher Mostly just because I don't think there's a >50% probability this would happen in 2028 rather than 2027 or 2029 or any other year.
I also don't think humanity could make use of this many LLMs per day. There could very likely be just a few popular LLMs that don't require any fine-tuning and can do any text-based task with just a prompt, or a single system with interchangeable modules, or we could be using a paradigm completely different from LLMs.
@CDBiddulph My guess is that LLMs will become so easy to produce that either the process of producing them will in essence be automated by a vast amount of them or that the interchangeable modules that already exist, in terms of certain forms of prompt engineering or fine-tuning become so diverse, that it takes no effort for humans to produce thousands and create an open market to now test what happens in regards to human behavioral economics to get an edge on technological trends bootstrapped from them in multi-faceted directions.