🏅 Top traders

What a shameful display from the employees who remain. They may have lost the last impactful opportunity to resign.

Not to mention they are very likely to face prison time now if we don't pause before something really bad happens. Maybe the ICC.

@Haiku what? How is this what they’re doing illegal lol? It’s just matrix multiplication

@Haiku I agree that we are increasingly heading into more and more dangerous waters, but I think that’s a bit too simplistic and extreme of a statement. There will likely be future opportunities under all but the most extreme and shortest of RSI/intelligence explosion timelines.

The second part about prison time is indeed something I think lab employees should consider themselves risking, though I doubt it’s likely at the moment rather than a marginal risk that’s growing at an unknown but fast pace. That is to say, I doubt a massive rogue AI catastrophe such as many billions in damages and lives lost from a cyberattack incident are likely in the next few months, conditioned on being attributable to Anthropic/OpenAI models directly. Passing a few months out everything becomes a lot less certain.

@JasonMendoza2008 That's what's known as the fallacy of composition. (Biological warfare is just molecules bumping into each other. Tigers are just chemical reactions. The human brain is just neurons spiking.) Everything that exists and every event that happens emerges from simple rules: the laws of physics. Many of those things can kill you, especially the things that can think and understand and reason, by every functional definition.

60% of published AI researchers believe that there is a double-digit chance of literal human extinction from AI. Most of the CEOs and lead scientists of the companies admit that the risk is real. Even if you disagree with the majority scientific view, the fact that the companies say that they might destroy the world and that they are going to go ahead anyway should concern everyone.

@DavidHiggs Indeed, I think there is a >50% chance of at least one more significant warning shot before it is too late to begin to intervene. I might put it at 85%; there is more room to be surprised than I would like.

@Haiku everything you said is true but has nothing to do with employees going to jail. You completely missed my point. As long as there aren’t any laws preventing me from doing matrix multiplication to the point that I can decimate humanity I shouldn’t be scared of going to jail. Justice is not retroactive.

@JasonMendoza2008 I disagree. If matrix multiplication of a certain type has sufficient risks that are foreseeable even if uncertain, then I support laws or court decisions retroactively holding such research liable including prison time. Not arbitrarily of course, there would need to be cross-referencing of existing precedent, consideration of precautions taken and negligence standards, etc.

But ultimately, we shouldn’t need to pass new laws to hold extreme risk taking legally accountable. Just like the Nuremberg Trials broke new ground rather than attempting to narrowly prosecute “crimes against humanity” according to the exact letter of existing laws, legal action to prevent catastrophic risk from AI should also break new ground including retroactively. The spirit and intention of existing legal frameworks should be adapted to this new frontier however is necessary to hold people accountable for playing Russian Roulette with the human race.

@DavidHiggs Agree to disagree!

@CDBiddulph does this count?
https://time.news/openai-safety-leader-david-robinson-resigns-and-blasts-corporate-culture/

@bobalobascrob if not please let me liquidate.

@bobalobascrob they resigned "last week" (week of sept. 21-27) which is in september

@bobalobascrob and work in safety, citing safety concerns and a blatant disregard of safety procedures

@bobalobascrob I bet up a lot if you want to liquidate with a profit without any risk

@JasonMendoza2008 nah i'm confident that this would resolve YES.
Your access has been blocked – The Atlantic

@bobalobascrob sure

@JasonMendoza2008 yeah basically i've liquidiated early one too many times and this time im gonna trust myself

i have never been wrong about anything large before

@bobalobascrob no need to explain yourself young man

ethical concern most apparent and similar to "loss of control": "Even short of a full loss of control, we could see autonomous swarms of AI agents that act without human permission. Imagine “rogue” agents that work like teams of hackers (for example, holding hospital computer systems for ransom) but never need to sleep."

@bobalobascrob Yes, looks right to me!

@CDBiddulph

I'll wait for two extra days, in case any resignations happen during September but are only publicized later

I could be missing something but doesn't the description seem to imply anything announced after oct 2 shouldn't count?
I bet thousands of NO on the option that resolved YES at Oct 2, 11:45pm. ig i totally get it if you decide to still resolve on later announcements but am i misreading this or is it kind of like shifting the vibe resolution criterion bc it matches the spirit better (which tbc is fine)?

@Bayesian Yeah, I guess it is kind of shifting the resolution criterion because it matches the spirit better. But if you only look at the title, the way I'm resolving it is more accurate to the title, if that helps. Point taken about the criteria change, but I'm not switching it now

@CDBiddulph im the one who suggested to extend. Apologies.

@CDBiddulph do these all resolve no if nothing comes out by market close? what if some new news comes out tomorrow

@bobalobascrob @CDBiddulph i think you should reopen for a week or two. People might be bounded by contracts during their notice period.

@JasonMendoza2008 Hm sure, I'm not in a rush