Related questions
@GensingGenis Thank you. To confirm, 3 days per week spent using AI, 4 days spent not using AI counts? And I assume taking a 1 week vacation a few times a year does not disqualify a person?
Proposal: >50% of work days? (Although you mentioned non-work usage so perhaps this is not a good criteria.)
Alternatively, is the question about the proportion of the US population which will be using AI on a given day? So if >169 million US citizens use AI on an average day, the market resolves YES?
(I apologies for nitpicking, but it is valuable for markets to have legible resolution requirements whenever possible.)
@GensingGenis What does "AI" mean in this context? Does any usage of a neural net count as using an AI?
@Amaryllis Here’s a good answer:
For something to be considered as "having AI", it needs to be able to deal with unknown environments/circumstances in order to achieve its objective/goal, and render knowledge in a manner that provides for new learning/information to be added easily. There are many different types of well defined knowledge representation methods, ranging from the popular neural net, through to probabilistic models like bayesian networks (belief networks) - but fundamentally actions by the system must be derived from whichever representation of knowledge you choose for it to be considered as AI.
https://ai.stackexchange.com/questions/1507/what-are-the-minimum-requirements-to-call-something-ai
@GensingGenis I apologise for arguing over semantics, however I continue to be unsure of the resolution criteria for this market.
What are we using the phrase "unknown environments/circumstances" to mean? Generalization outside of the training distribution?
What are we using the term "goal" to mean? I assume that "minimize the loss function" does not count as a goal in this context?
According to the given definition, repeatedly sampling a multi-layer neural net trained to turn a string of text into a probability distribution over the next token would probably not count as AI given that it does not have a "goal" in a meaningful sense? (Does, "accurately predict the next token" count as a goal?) (Does it change if the model has had RLHF applied to it?)
On the other hand, a recommender system does arguably have a "goal", and thus would count? In which case ~anyone who uses
Can you give some examples of current things which count as "using AI"? For example:
1) Interactive use of Chat GPT
2) Sampling text strings from a pure self supervised language model
3) using speech to text system to transcribe voice
4) using a diffusion based image generation to generate an image
5) using Google search (for basic website engine search and ranking capability, not whatever LLM based summary tools they are adding)
@GensingGenis Do you refer to the answers to the ai.stackexchange.com question? I have read all 5 answers. My questions are not answered.
As best I can tell, according to the definitions given, a speech to text RNN or a recommender system arguably counts as "AI". They observe the world, and have an internal representation of their knowledge, and use it to pursue "goals" of a sort. Do you intend to resolve based in this interpretation of the definition?
(I am reluctant to put more mana into this market until I have clarity on resolution criteria.)
@Amaryllis Can’t answer this today, if anyone finds this confusing please let me know and I’ll clarify in the coming days
@GensingGenis Would searching on Google count? Especially if they expand their generative AI answer features
@CelebratedWhale If it’s a glorified google search I would have to say no. If it’s an AI agent like a Chat GPT meets Siri that can act on its own then yes.