Oh Hi

@nosilverv This is a feature of humans, not a bug. It encourages cooperation which is beneficial overall. That’s the real rationality

Ultimatum Game: experimenter gives you 10$. But you only keep it if your confederate accepts how you suggest to split it. “Rationally” they should accept anything non-zero: it’s free money. But irl people often reject anything below 70:30 and the most frequent offer is 50:50(!)

@Glhuun I would probably not choose to risk dying to reduce the risk of others dying. Might be selfish, but I weigh my own life more

@ThroughRate Honest question cuz fucked if I know the answer lol

@cremieuxrecueil @rnovak1988 Mutual rudeness there. Anyway, your wording was confusing on first pass. You wrote "and easily 1-in-6 to 1-in-5 hit blue by mistake" when I understand you meant "then easily [...]"

@rnovak1988 Hey, you might be retarded, and that's OK! I'd still press blue to save you. But try to understand that the 6% is being ADDED ONTO the rate for kids. It's clear from my wording, but it's understandable you'd miss this if you have a learning disability.

Accurate cartoon. In my polling, almost 6% of people said they picked the wrong button when they were prompted about it. Assume the global error rate is higher and kids' responses are random, and easily 1-in-6 to 1-in-5 hit blue by mistake. So, if red wins, society is ruined.

@JamesCanaryII @Jebus Nowhere does this plot returns. The y-axis is skill and the x-axis is percentile. If you say skill=returns then it is EXPONENTIAL returns, literally the exact opposite of diminishing returns.

@Jebus Its the law of diminishing returns

This is the actually life’s standard skill distribution chart

@Jebus Wrong. The word "skill" is not defined. If it represents measurable performance then many domains follow a diminishing pattern, not exponential. Example: 100m sprint. Going from 50th to 99th percentile is like 15s to 11s. Going 1s faster takes puts you in 99.9999th percentile

@Lathalian @shinkanko @Jebus The problem is that "skill" is not defined. If it represents measurable performance then many domains follow a diminishing pattern, not exponential. Example: 100m sprint. Going from 50th to 99th percentile is like 15s to 11s. Going 1s faster takes puts you in 99.9999th percentile

@shinkanko @Jebus That's not what the chart shows - the median guy on this chart still has 2x more "skill" than the guys before him, which is clearly appreciable in many disciplines. You can't understand how much better the top end is until you get "close" and see how far you still have to go.

@PhilippeLaban Incredible paper. The backtranslation loop idea is so elegant. Wish I came up with that

Conclusion: Current LLMs are unreliable delegates. They introduce sparse but severe errors that compound and corrupt work documents. We release DELEGATE-52 to encourage the study of delegated work across knowledge work domains: 💻 github.com/microsoft/DELE… 🤗 hf.co/datasets/micro… 📄 arxiv.org/abs/2604.15597 Work done with my wonderful colleagues Tobias Schnabel and @ProfJenNeville at @MSFTResearch

New paper! LLMs Corrupt Your Documents When You Delegate LLMs are enabling a new way of working: delegated work, where users supervise an LLM as it edits documents on their behalf. Delegation requires trust: does the LLM complete tasks without introducing errors? We simulate delegation across 52 professional domains and find that LLMs Corrupt Your Documents When You Delegate. 🧵1/N

@itsssthulsss Julian Assange did exactly that and hid in an embassy for 7 YEARS to avoid being arrested

guys i have a question. if you commit a crime abroad and go hide in your country’s embassy… can they actually not arrest you??

@Ray_Mordt @grok @pmarca How do you reject the argument when you haven’t even read it? If you value critical reasoning, then this is precisely *why* you should read things you may disagree with

@grok @pmarca Thanks, I will not bother reading it. I reject it's core argument, and therefore @pmarca 's assertion here.

If you want the scientific demolition of introspection, this is the book: amazon.com/Mind-Flat-Rema…

@AlexHormozi Did you have your biceps out?

I got carded over Thanksgiving break with family buying beer. So yea, that was peak. Small victories.