Can you please clarify what you mean by "directly influences strains being created?"

Do you mean 'strain' of a species of pathogen, or do you mean 'species of pathogen'? Sometimes people aren't clear here.

What do you mean by 'created'? Combination of existing natural virus components, direct synthesis of RNA/DNA used to make viral proteins, creation of a whole virus from scratch, etc.

GOF isn't necessarily anything to do with dangerous functions. Do you include functions that are not directly relevant to the risk presented by the pathogen to other organisms?

What do you mean by 'influences'? Do you mean use of an AI can be shown to have contributed to the decisions of a particular researcher? Do you mean that an AI is facilitating the broader work of a lab which influences which strains the lab is able to create or does create? Or that the AI itself is directly influencing the work stream of a researcher/lab to include a particular pathogen/strain? Or that the AI itself is doing the strategy, research, and 'creation' of a pathogen?

I think this question is quite complicated and vague and without clear definitions this will not be a useful market.

@WXTJ i mean any situation where the AI has a significant ability to influence the DNA/RNA of a bacteria or virus being created, all examples you gave would count for YES

If you have any ideas on how to improve the criteria I'd be happy to hear them, I think the general idea of "will somebody let the shoggoth design pathogens on purpose" is pretty clear

@CodeandSolder Thank you.

I think in that case the resolution of this market might involve a lot of debate reacting to specific events, so predicting how that debate would go will be a major factor in people's predictions, on top of their predictions about the world.

e.g. as you say "AI is facilitating the broader work of a lab which influences which strains the lab is able to create or does create?" would count, any lab using a narrow AI to predict protein folding based on sequences might meet the criteria. This is because this could lead to them working on manipulating functions of a virus that weren't available to them before, creating new functions they couldn't have predicted before, or just doing much more work as a result of the improved capabilities, leading them to work on projects they wouldn't have without the help of the AI model.

This is a very different type of scenario to what some people might think reading your title, and is perhaps more likely in the near-term. In my opinion it is also a very risky situation, comparable to that of an AI directly designing novel viruses. But now that you have clarified your criteria, it seems that this market is about a broad and varied set of events. That is fine, it's just less useful to bet on and debate.

@WXTJ I absolutely see the problem you describe, for example using something like AlphaFold does not fulfill the spirit of the market IMO and it's not clear
Like I said I'd love to hear any suggestions on how to make this market more useful

@CodeandSolder Thank you for engaging with my pedantry :)

To be honest, on reflection, I'm not sure if it's a good idea to put out specific practical scenarios here. Perhaps it's best to keep things vague. Despite my above comments I have to admit I don't have any good ideas of how to make it specific/precise, useful as a prediction, and not provide some degree of attention/information-hazard. So for taking up your time with that I apologise! Thank you at least for the opportunity to think about this.

@WXTJ always happy to talk.

I doubt they aren't doing the most stupid, most obvious idea in this set in a lab somewhere TBH

I don't want to overconstrain the market too much though, just in case they figure out a new way to be irresponsible.