Thank you for all your energy this year Ben! Interestingly, I've been thinking myself about those parallels between railway and internet infrastructure builds recently, and their value post-bubble; and wondering whether an AI bubble pop could offer us a similar silver lining. It led me to wonder what 'tracks' we were actually laying down here, and your newsletter intersects with a number of them... For example, even ignoring the "neocloud" risks, I wondered how a typical investor - someone whose ideology is essentially a belief in "growing the size of the overall economic pie" (vs., for example, redistributing it) - can justify a (conservatively-priced) US$30,000 H100 with the equivalent energy-use of a US citizen and a useful working lifespan of 1-3 years, compared to, say, a life-changing US$30,000 'investment' (education, social support, business loans etc) in a human with a working lifespan of 40-60 years? We might start to see more shape to this in 2025, but I'm still struggling to find optimism among the foundations that the ceaseless cycles of 'more compute' will leave us with, if this bubble does burst... Anyway, 2025 will undoubtably be interesting either way! Have an excellent Christmas and New Year, and thanks again.
I think that you frame the issue very well as a choice of investment values - leading to the perennial question what is the objective function of humanity (or at least the economy...?)
From a human near-term perspective it's clearly a values question: concentrated material wealth (/status/dopamine) over wider measurements of "wellbeing" for humanity and the environment on aggregate.
But increasingly I think about it from a more external medium-term future perspective "how would the superintelligent AIs (plural) look at humanity?" Like competitors to be fought (Terminator)? Pests to be exterminated? Exotic wildlife to be preserved or largely left to its own devices? (If so, is the relative gap like humans to monkeys, insects, bacteria, plants...) Or symbiotic lifeforms to be nurtured?
Let the investors throw their money around, it's unlikely to change the direction of evolution.
Thanks for all your support throughout 2024, Tim, looking forward to a wild ride in 2025!
Haha, there's probably some relevant meaning to be found in how I've found myself so bogged down with 'regular' work stuff that my own writing has been badly neglected... I guess we're all (investors included) burdened by what we are compelled to prioritise 😉
Great work Ben. Skating on ice off unknown thickness over water of unknown depth and there-be-monsters.
Meanwhile
https://www.theregister.com/2024/12/10/ai_slop_bug_reports/
Cheers John! Crazy times about to get crazier... thanks for all your support and engagement during 2024... enjoy your Xmas break!
Thanks Ben. Have a great holiday, then continue the good work!
Thank you for all your energy this year Ben! Interestingly, I've been thinking myself about those parallels between railway and internet infrastructure builds recently, and their value post-bubble; and wondering whether an AI bubble pop could offer us a similar silver lining. It led me to wonder what 'tracks' we were actually laying down here, and your newsletter intersects with a number of them... For example, even ignoring the "neocloud" risks, I wondered how a typical investor - someone whose ideology is essentially a belief in "growing the size of the overall economic pie" (vs., for example, redistributing it) - can justify a (conservatively-priced) US$30,000 H100 with the equivalent energy-use of a US citizen and a useful working lifespan of 1-3 years, compared to, say, a life-changing US$30,000 'investment' (education, social support, business loans etc) in a human with a working lifespan of 40-60 years? We might start to see more shape to this in 2025, but I'm still struggling to find optimism among the foundations that the ceaseless cycles of 'more compute' will leave us with, if this bubble does burst... Anyway, 2025 will undoubtably be interesting either way! Have an excellent Christmas and New Year, and thanks again.
I think that you frame the issue very well as a choice of investment values - leading to the perennial question what is the objective function of humanity (or at least the economy...?)
From a human near-term perspective it's clearly a values question: concentrated material wealth (/status/dopamine) over wider measurements of "wellbeing" for humanity and the environment on aggregate.
But increasingly I think about it from a more external medium-term future perspective "how would the superintelligent AIs (plural) look at humanity?" Like competitors to be fought (Terminator)? Pests to be exterminated? Exotic wildlife to be preserved or largely left to its own devices? (If so, is the relative gap like humans to monkeys, insects, bacteria, plants...) Or symbiotic lifeforms to be nurtured?
Let the investors throw their money around, it's unlikely to change the direction of evolution.
Thanks for all your support throughout 2024, Tim, looking forward to a wild ride in 2025!
Haha, there's probably some relevant meaning to be found in how I've found myself so bogged down with 'regular' work stuff that my own writing has been badly neglected... I guess we're all (investors included) burdened by what we are compelled to prioritise 😉