@rodrigotellom Very fun!
Notes
- Just found and binged this blog. Adam has a brilliant, fresh, and at times hilarious style. He also narrates his posts (always in a single take), and he's a very entertaining narrator https://x.com/a_m_mastroianni/status/1477405147319971845
- @benjiwheeler Fun!
- RT @byjonathanleung: I'm new to the worlds of Doug Engelbart, Ted Nelson, and Alan Kay and want to draw inspiration from their work to reth…
Reposted by Steve
Text may be partial · Read original
- The story of AI advancing is not us all waking up one day to have veritable gods in our lives, but for us ever so slowly to have smarter and smarter systems
- Most interestingly are the weird states: ie X% of the world seems stupider than AI and Y% seems smarter X=10, Y=90 X=50, Y=50 X=90, Y=10 X=99, Y=1 X=99.999, Y=.001 Makes you think, huh?
- As a simplistic measure, we could say AI now has an IQ of 70. So I realistically couldn't tell the difference between it and a human of IQ 70, modulo a few inconsistencies that I might be able to suss out if I am persistent and savvy enough
- Sidebar: Could we could actually measure GTP3's IQ? I know people gave it SAT questions...
Text may be partial · Read original
- Anyways, the point I'm getting to is that as AI systems get smarter, the Turing Test becomes not "is this a computer or a human" but "is this entity smarter than the smartest AI system of today"?
- Like @paulg says: to tell if someone is smart, lob a couple over the net and see if they can hit em back If you put a genius behind a chat interface, I can tell easily it's not an AI If you put a dumb human behind a chat interface, I might not be able to tell
- As I think about the Turing Test and ChatGTP, I think about how if you showed me ChatGTP 5 years ago, I think it maybe could've fooled me it was a human But now I know what AI systems can do, so I know where the bar is https://x.com/stevekrouse/status/1602805545882329089
- @malonehedges @jarredsumner I was about to make this joke
- RT @swyx: @stevekrouse https://x.com/minimaxir/status/1600258754993262594?s=46&t=Q1KnnOaUeENMGRuBabhGYA
Reposted by Steve
- @swyx Has ChatGTP actually passed the Turing Test? I feel like I could tell if it were a human or computer fairly easily
- learning to surf hype cycles is one of the hardest parts of being in tech
- @pquiggles Ha! Great point! I used the wrong metaphor. It's not like talking to a stupid person. (Although that metaphor still holds when we get there.) It's more like talking to siri that's 2x better. Still haven't found a use for it that would make me upset if it went away
- Caveat: I love GitHub copilot. So I could see something like this working as a writing assistant (Lex). But the chat interface isn't blowing me away like it is others
- Honestly though, I don't really get the hype. It's sorta kinda neat sometimes if I work really hard at prompt engineering. It's like talking to a really stupid person that is overconfident, wordy, and knows a bit about everything. Why would I want to talk to such a person?
- @ftlsid still rationalism you're looking for a version of rationalism for people who have social skills
- Finally found a use for this AI thing https://x.com/stevekrouse/status/1602717873675423744/photo/1
- @NYT_first_said Someone should submit the typo!
- Any of my friends have @tana_inc invites? https://x.com/benjiwheeler/status/1602519071076982784
- RT @goblinodds: actively looking for junior dev positions rn, hmu if you have leads or know recruiters i can talk to! -self-taught, fronten…
Reposted by Steve
Text may be partial · Read original
- RT @glideapps: @glideapps was designed to help anyone build apps This included non-technical operations managers and technical power users…
Reposted by Steve
Text may be partial · Read original
- But that won't stop me from reading @slatestarcodex and @ESYudkowsky, and taking all the wonderful ideas that EA has to offer
- To be clear, this thread has almost nothing to do with SBF This thread is more in response to all the people now saying "EA is awful" and others saying "EA is clearly the best" As an EA-adjacent, I thought my perspective might be helpful
- There so much good in EA and its ideas. In fact I think that the ideas are so good that it attracts a certain kind of person that loves to have moral certainty and superiority. There's just too high a % of those folks for me to feel comfortable getting on the bus
- Obligatory disqualifications: this is not representative of the whole community blah blah. That's definitely true. I know lots of wonderful EA people that seem to pose zero risk of becoming supervillains
- However I will say that for me personally, every single time I try to wade a tiny bit into the EA community, I run into this religious or morally-certain strain of thinking that pushes me back out
- Sam Harris's recent podcast with Erik Hoel (#305 Moral Knowledge) has helped me put this aversion into words Playing moneyball with charities: obviously good Giving more and spending less on luxuries: great idea
- The line is when people extrapolate to unintuitive conclusions and want to take them seriously at scale. This is where fanaticism starts, and it's worth pointing out and pushing back against
- They later said that if born 100 years earlier they definitely would've been a communist. They said that they're one of the people that wants to be right, do good, and isn't scared to take moral principles to their unintuitive conclusions, at scale Is there anything scarier?
- I once had a chat with an EA person who spent the whole time arguing passionately that everyone (including me) should drop everything to bring about the AI singularity as soon and positively as possible, so we can all cede responsibility to it
- I even said on the call, "don't you think you sound like a religious person telling us how to bring about rapture, the end of days, where God descends onto earth to judge us?"
- People joke that EA is a cult or religion. And yes it is somewhat of a joke, but cultishness and religiousness are best understood as spectrums not binaries, and there are times when EA folks exhibit a bit too many cultish/religious characteristics
- I think this tweet aged well 😅 As someone who loves EA thinking but has never taken the plunge into the community, I just wanted to say a few words about EA https://x.com/stevekrouse/status/1581353649334542341
- Actually it seems like there's an email-based fallback: 1. If you fail to login to your account on your phone, it will email you a 6-digit code you can use to login 2. After logging in on your phone, it will allow you to setup (a second?) passkey for that account
- The password is dead. Long live the passkey! This was a sexy demo! 1. Sign up on your phone with your phone lock as your password 2. If you try to login on desktop, it uses bluetooth to connect to your phone so you can use your phone lock to login https://x.com/maccaw/status/1601879505592733697
- The experience in the other direction (signup with computer first, then login with phone) wasn't ideal My andriod didn't seem to know how to ask my mac for the passkey (like chrome on my mac knew how to ask my andriod for its passkey)
- @johncutlefish @tikhonjelvis Your analogy explains why it's so mind-boggling! After roads are built, they effectively fire everyone who built it. And that's fine right? So why should it be different for tech infra? Just because 14k people built it, doesn't mean 14k people need to maintain it or grow it


