
Inspired by "I am Bing, and I am evil". See also What an actually pessimistic containment strategy looks like.
A notable attack could include a bombing, an attack on someone who works there, a hack, etc. The attack does not need to be successful; a bomb that doesn't detonate would still count. However the planned attack must actually matter. Some graffiti on the wall of an AI lab is not going to resolve this YES, though more significant vandalism like cutting the internet cables could count.
A large, sustained, in-person protest would also count, even if nothing is violent. Must be at least 500 people over the course of at least a week, and actually cause disruption to them.
Any hack counts if it's seriously disruptive. A data breach alone is not good enough.
🏅 Top traders
# | Name | Total profit |
---|---|---|
1 | Ṁ1,908 | |
2 | Ṁ1,152 | |
3 | Ṁ531 | |
4 | Ṁ502 | |
5 | Ṁ465 |
People are also trading
@IsaacKing What about a data breach or otherwise releasing information with the intent to hurt a company’s credibility? Like revealing a private chat or an internal memo
@duck_master hmm, it's not obvious to be what the arbitrage is here. There's a big difference between a protest (which this market includes) and a terrorist attack (which seems to be what the other market requires) for example.
Writers guild protesting the use of chatgpt in a strike https://www.euronews.com/culture/2023/05/03/why-the-ai-demand-from-the-writers-guild-strikes-is-the-most-important-talking-point
@Boojum good for context, but I'll note that there's no "leading AI organization" being targeted here.
@TeddyWeverka Seems like what matters is whether a week is "sustained" and whether how ever many people show up is "large," since it is clearly an in-person protest.
@IsaacKing any clarification on what it would take for this sort of protest to count? Here's the summary, "📢 Hey everyone! Im organizing a week-long picket around OpenAI's HQ in San Francisco, mid May (So in 2 weeks!), to call for 6 month pause on AI development, to get public attention. Stay tuned for the updates here. Join me, spread the word! Together, we will succeed! #PauseAI"
@JamesRomeril I'd argue that would depend on the impact of the hack - data or model exfiltration NO, something like ransomware or DDOS YES
@PipFoweraker IMHO the intent is still part of that. E.g. model exfiltration could be motivated by a competing company's greed, but data exfiltration could be "Look how much data this chatbot knows about you!". It's not clear if that's quite linked "enough" to AI itself, but it kinda points toward what that could look like.
@DylanSlagh Hmm. If they're doing capabilities research, I think that should count, yeah. If they just use other people's AIs to make and sell art, no.
@IsaacKing Midjourney trains their models these days fyi, e.g. Midjourney v4 was trained with TPUs
https://cloud.google.com/blog/topics/systems/tpu-v4-enables-performance-energy-and-co2e-efficiency-gains