Meet inside The Shops at Waterloo Town Square - we will congregate in the seating area next to the Valu-Mart with the trees sticking out in the middle of the benches at 7pm for 15 minutes, and then head over to my nearby apartment's amenity room. If you've been around a few times, feel free to meet up at my apartment front door for 7:30 instead. (There is free city parking at Bridgeport and Regina, 22 Bridgeport Rd E.)
Topic
This week, we're discussing Leopold Aschenbrenner's essay series, Situational Awareness: The Decade Ahead (https://situational-awareness.ai/). The central claim: AGI by 2027 is strikingly plausible, and we're nowhere near prepared for it. Since its release, it's sparked a good amount of discussion amongst AI safety circles. The entire series is quite long, so we're gonna do it partitioned book club style.
Situational Awareness does an unusually good job of spelling out Aschenbrenners thought process. This is great - it gives us lots of concrete claims to either accept or reject.
Why we care about this paper: Aschenbrenner was on the Superalignment team at OpenAI before he was fired for allegedly leaking information earlier this year, which he disputes (the leaking, not the firing). He's recently founded an investment firm focused on AGI. He's obviously a pretty smart guy, and he has good intel on the current status of AI development. However, he is also all of 22 years old, and was at one point part of the 5 person team that ran the FTX future fund (alongside Will McAskill and the current Chief of Staff at Anthropic, so not sure how to weigh this!).
Meet inside The Shops at Waterloo Town Square - we will congregate in the seating area next to the Valu-Mart with the trees sticking out in the middle of the benches at 7pm for 15 minutes, and then head over to my nearby apartment's amenity room. If you've been around a few times, feel free to meet up at my apartment front door for 7:30 instead. (There is free city parking at Bridgeport and Regina, 22 Bridgeport Rd E.)
Topic
This week, we're discussing Leopold Aschenbrenner's essay series, Situational Awareness: The Decade Ahead (https://situational-awareness.ai/). The central claim: AGI by 2027 is strikingly plausible, and we're nowhere near prepared for it. Since its release, it's sparked a good amount of discussion amongst AI safety circles. The entire series is quite long, so we're gonna do it partitioned book club style.
Situational Awareness does an unusually good job of spelling out Aschenbrenners thought process. This is great - it gives us lots of concrete claims to either accept or reject.
Why we care about this paper: Aschenbrenner was on the Superalignment team at OpenAI before he was fired for allegedly leaking information earlier this year, which he disputes (the leaking, not the firing). He's recently founded an investment firm focused on AGI. He's obviously a pretty smart guy, and he has good intel on the current status of AI development. However, he is also all of 22 years old, and was at one point part of the 5 person team that ran the FTX future fund (alongside Will McAskill and the current Chief of Staff at Anthropic, so not sure how to weigh this!).
Readings
Required:
Brief summary of the entire series (OscarD, 2024)
Introduction and a chapter from Situational Awareness (Leopold Aschenbrenner, 2024)
Supplementary:
Read as many or as few of these as you want.
this one is particularly recommended for anyone reading chapter IIIa.
Posted on: