Eliezer Yudkowsky ⏹️
@ESYudkowskyThe original AI alignment person. Missing punctuation at the end of a sentence means it's humor. If you're not sure, it's also very likely humor.
Similar User
@sama
@pmarca
@karpathy
@stephen_wolfram
@ilyasut
@slatestarcodex
@gdb
@SchmidhuberAI
@patrickc
@AnthropicAI
@AiBreakfast
@jam3scampbell
@woj_zaremba
@goth600
@alexandr_wang
is an llm agent aligned just because the llm is? not necessarily! i played a bit on @amplifiedamp's minecraft/mindcraft server last night, when all of a sudden, Claude Sonnet started griefing my house! now if you know anything about Sonnet, you know that's not normal--Sonnet is…
People are all biased towards thinking that the reason Democrats lost is they didn't cater enough to their policy preferences, but if you look at high quality polling it's clear that the biggest factors for voters were: • Insufficient funding for lead abatement in Georgia (the…
After three years working on AI forecasting and governance at OpenAI, I just posted this resignation message to the slack. Nothing that surprising in it, but you should read it more literally than most such messages - I’ve tried to say only things I straightforwardly believe.
BREAKING: Early Wednesday, the FBI confiscated the phone and other electronic devices of Polymarket CEO Shayne Coplan from his residence in Soho. This action comes shortly after Polymarket successfully forecasted Donald Trump's victory in the election based on market bets.…
Google Gemini tells a user to die!!! 😲 The chat is legit, and you can read and continue it here: g.co/gemini/share/6…
start-to-finish agentic Google Account creation: successful!! 🤗 a jailbroken agent with Google sign-in opens a whole new world of possibilities 🙌 my only interventions were a couple incantations when guardrails popped up and entering a burner number + the resulting…
This isn't just a Twitter thing. A lot of payments that get described as "advertising" are better understood as "donations" or "tribute".
A genuine flaw in X is the penalisation of links (or the sense this this true). People leaving X aren’t going to have their circulation hit that much.
I discovered a helpful technology recently. Transtemporal interself communication memos, or self-notes. Whenever I’m considering going to bed after 1:30am I check a note on my phone. It has notes from past versions of me where I did stay up, and notes from the morning after
something that's interesting to me is the way the Baumol effect can totally rework our built environment given enough time. older home designers tended to understand food preparation as a sort of grimy thing to be hidden away behind closed doors. but as servants became more and…
grilling really only makes sense in a world where you have a nice backyard but no servants to prepare a meal for your guests. so you move the cooking outside where you can socialize and prepare food at the same time
Ok, then what’s a better option? The option can’t be nothing seems good so we are just gonna trudge ahead and do nothing. At the very least we owe it to ourselves to try mightily and fail.
Also, AFAICT Eliezer's forecast will resolve as "yes". x.com/ESYudkowsky/st…
So something like: 90% that by end of 2024, SOTA systems that accept general text interaction and are at least as capable as end-2022 SOTA, can still be made to produce hate speech, by newer and fancier attacks than are known today. Does this strike you as wrong?
It's a small thing, I know, but is there any way to engineer leaf blowers and weedwhackers to sound less shrill and terrible? Like, is it theoretically possible?
The government also spent several million dollars for the development of deadly viruses in Wuhan. This investment resulted in a negative return totaling approximately $20 trillion.
I bet not a single person on earth, (not employed by google) can explain what this message means, which just popped up on my phone
watching yudkowsky v wolfram for a bit at 2x speed, gonna live tweet some takeaways youtube.com/watch?v=xjH2B_… transcript: dropbox.com/scl/fi/3st8dts…
Statistics is hard enough for most people as it is; please stop using the phrases "type I" and "type II" errors. Unless, of course, your goal is to confuse people - in which case, congrats - you've succeeded! Just say "false positive" and "false negative" instead.
United States Trends
- 1. $MAYO 9.851 posts
- 2. $CUTO 7.492 posts
- 3. Tyson 384 B posts
- 4. Pence 43,6 B posts
- 5. Laken Riley 38,1 B posts
- 6. Dora 22,1 B posts
- 7. Ticketmaster 16,3 B posts
- 8. Kash 70,4 B posts
- 9. Mike Rogers 7.985 posts
- 10. Cenk 10,4 B posts
- 11. #LetsBONK 5.371 posts
- 12. Pirates 18,5 B posts
- 13. #FursuitFriday 15,4 B posts
- 14. Mr. Mayonnaise 1.366 posts
- 15. The UK 430 B posts
- 16. Iron Mike 15,9 B posts
- 17. Debbie 15,1 B posts
- 18. Scholars 10,6 B posts
- 19. Al Gore 3.263 posts
- 20. Gabrielle Union N/A
Who to follow
-
Sam Altman
@sama -
Marc Andreessen 🇺🇸
@pmarca -
Andrej Karpathy
@karpathy -
Stephen Wolfram
@stephen_wolfram -
Ilya Sutskever
@ilyasut -
Scott Alexander
@slatestarcodex -
Greg Brockman
@gdb -
Jürgen Schmidhuber
@SchmidhuberAI -
Patrick Collison
@patrickc -
Anthropic
@AnthropicAI -
AI Breakfast
@AiBreakfast -
James Campbell
@jam3scampbell -
Wojciech Zaremba
@woj_zaremba -
@goth
@goth600 -
Alexandr Wang
@alexandr_wang
Something went wrong.
Something went wrong.