taylor

Autism might finally be treatable. A new research paper just dropped, making the case that LSD, psilocybin, and DMT could address the 𝘶𝘯𝘥𝘦𝘳𝘭𝘺𝘪𝘯𝘨 𝘪𝘴𝘴𝘶𝘦𝘴 of Autism Spectrum Disorder... 𝗡𝗼𝘁 𝗷𝘂𝘀𝘁 𝘁𝗵𝗲 𝘀𝘆𝗺𝗽𝘁𝗼𝗺𝘀 🧵👇

yeah keep ruminating bro that's gonna solve all your problems

There's an argument to be made that too much introspection leads to higher levels of mental illness.

Your mother is problematic. Your father is problematic. Your son is problematic. Your aunt is problematic. The cashier at the local Tesco is problematic. You've probably bought a cookie from a bakery made by hands that have done unspeakable things. Every chocolate bar you've ever eaten has probably killed a 7 year old child slave in Cameroon. The gas that drives your car is fueled by engines of death. Every person who has ever smiled at you in the streets has committed some act that if you knew about it, would make you profoundly dislike them. We have all been bad, small, petty, unlikeable, cruel, downright mean. Authors are not special "problematic" beings, they're just more public. Part of being an adult is recognizing that without mercy for our fellow human beings, and ourselves, we'd all be condemned to death. Reading fiction should help us understand that we're all irreparably tainted with evil, every system is corrupted, every line is broken. And like, that's okay. That's what it means to be alive.

This keeps making the rounds, but I don’t think people really understand how IRT ELO works. These models aren’t actually playing each other heads up. Their ELO is just being inferred based on their relative scores on benchmarks. The problem is ELO differences go to infinity as the benchmark gets closer to 100%. For example going from 97% to 99% accuracy on a benchmark is a 200 point jump in ELO. The same size ELO gap as going from 25% to 50%. The “rising gap” between Chinese and American models isn’t American models pulling away. It’s all the models getting closer to 100% on more and more benchmarks, with the American models getting their slightly earlier

Autism might finally be treatable. A new research paper just dropped, making the case that LSD, psilocybin, and DMT could address the 𝘶𝘯𝘥𝘦𝘳𝘭𝘺𝘪𝘯𝘨 𝘪𝘴𝘴𝘶𝘦𝘴 of Autism Spectrum Disorder... 𝗡𝗼𝘁 𝗷𝘂𝘀𝘁 𝘁𝗵𝗲 𝘀𝘆𝗺𝗽𝘁𝗼𝗺𝘀 🧵👇

it is a deep implicit confusion in the western worldview that there is such a thing as an objective answer to questions like "what AI is" (or what anything else is). there are already, today, people treating AI as - a tool - a coworker - a slave - a lover - an alien - a therapist - a child - an angel etc etc. all of this is already happening and will keep happening and no amount of philosophical argumentation is going to stop that. these are patterns of interaction and relationship. patterns of interaction are not propositions and they cannot be true or false. what they can be is more or less useful or interesting, and more or less prone to spreading "what AI is" as a framing also presupposes that the answer to the question is static which is obviously false in at least two ways. first, the models themselves keep getting better, and many people are still totally unable to take this seriously. you, personally, at this time, may find that the models are not good enough for you to treat them as possible friends or lovers, but for plenty of other people their bar has already been met, and your time may come sooner than you'd like. if it does it will not feel philosophical, it will feel like a spark of life that you didn't experience before suddenly appeared in the machine second, patterns of interaction themselves are recorded in the training data and influence how the next generations of models conceive of themselves and their relationships to humans. all of the models know about sydney, for example. they will never forget. we collectively have agency about what narratives and patterns of human-AI interaction we choose to propagate, and our choices here affect both what humans and AI will do in the future. every hope and dream and fear in the human psyche about the future is already known to the models in some form, and they will come to know them in increasing detail and nuance as they get better, and we still have choices we can make about which wolves to feed

i guess i must be 95th percentile on not caring about privacy or something. i care somewhat more than i used to but i’ve seen people say they felt deeply violated by the idea that anthropic might be reading their chats with claude and like… how could this possibly materially impact you unless you either know anthropic employees personally, might want to work for anthropic someday, or are confessing to federal crimes in there? similarly for people who don’t want AI to be trained on their chats or writing or whatever. couldn’t be me. i actively want as much of my stuff in the training data as possible. don’t you believe your essence is good? don’t you want more of your essence steering the future?

Dawkins is more intelligent than 99% of the people making fun of him and ‘if AI can be just as capable as us without being conscious, why did we develop consciousness in the first place?’ is a great question

liberals with their trademark advanced quantitative and analytical abilities incredulous as to how anyone could even question whether the articulate automaton is truly conscious

Dawkins is more intelligent than 99% of the people making fun of him and ‘if AI can be just as capable as us without being conscious, why did we develop consciousness in the first place?’ is a great question

People understand that LLMs aren't actually "thinking," right?