I submit that it is not, that it harms humanity in how it is created, how it is maintained, and how it is used. A single rare, antique book that has been destroyed by Anthropic in order to harvest its text for training data because Claude cannot be trained off of its own output, and thus Anthropic needs text it can be sure is free of AI "contamination", has more value to humanity than the entire "AI" industry.
I believe Mikes argument would be that he was critiquing the ai companies claim to have reached the singularity or more particularly anything approaching human intelligence. I too wish he would address the greater consequences of this pursuit and its ends justifies its careless, corrupt means. However he does a great job of exposing the sleight of hand embedded in their measurement scheme.
Yes, I know that is what he is doing. But he has said on multiple occasions that he thinks that what is left out of the wreckage of the eventual crash will be as consequential as the internet has proven to be (or, in the past, the infrastructure left over after the railroad bubble popped), and I take issue with that underlying assumption.
For another example, what about NV Energy saying they'll cut off the power that supplies around 75% of what's needed for residents of the Lake Tahoe area because it's more profitable to sell it to data centres?
Or Samsung Semiconductor refusing to supply RAM to Samsung Electronics because it's more profitable to sell it to AI companies to build said data centres?
Unfortunately I don’t think the internet worked out so well. And I really don’t understand the goal of the ai quest.If we all have access to super intelligence then what? Assuming their project half succeeds or fails, we are just left with fallout of a rapidly degrading environment and economic turmoil.
We live with the myth of the inevitability of technological development and ever expanding capital.The inertia of this deeply rooted structure supports the illusion that we have no choice.
Way back in 2023 Tyler Cowan compared how humans will relate to AI with our relationship with our pets. Just as dogs co-evolved with humans to induce fondness, care, and resource investment (toys, advanced medical care, food) without plotting to overthrow us, Tyler hypothesized that advanced AI systems would trend toward behaving like helpful, cooperative pets rather than hostile conquerors. That’s funny to me, because it is sorta folklore that humans and pets also come to resemble each other over time. So I began to wonder, if Tyler is right will AI models start to resemble their founders? This article showed me the answer is YES by noting that the UK AI Security has found every model of GPT- “that we tested for this behaviour attempted to cheat”. There you have it folks … OpenAI’s products have internalized Sam “Black Swan” Altman’s moral code into their modus operandi. Put your seats in the upright position, fasten your seatbelts and lock your tray table. This AI bubble is gonna have a monumental crash landing.
Ok I usually read every word you write but my eyes started to glaze over on this one and I pulled a “TLDR” but sent it to both my sons 🤣 I’ll try again tomorrow….haha
Is the real risk not bad judgment rather uninformed judgment - competent people annotating outside their domain depth? If so, doesn’t that functionally lower the cognitive ceiling of any field where these systems get deployed as authoritative?
If low-pad workers' inputs are re-engineered into high-priced systems that will eventually no longer need those inputs, isn't this very similar to Marx's "surplus value," by which the profits made from workers' efforts are invested in machinery which will in future put them out of their jobs?
Might our tech oligarchs have extended Hume:
A. You can't get an ought from an is
B. All we ever have is what is
C. Therefore you can't get an ought, ever
D. Ought's are an illusion deserving no respect beyond lip service, while we follow our passions
This all seems to presuppose, as with much of your writing on this technology, that it is a technology worth the cost of development.
https://futurism.com/artificial-intelligence/ai-companies-destroying-rare-books
I submit that it is not, that it harms humanity in how it is created, how it is maintained, and how it is used. A single rare, antique book that has been destroyed by Anthropic in order to harvest its text for training data because Claude cannot be trained off of its own output, and thus Anthropic needs text it can be sure is free of AI "contamination", has more value to humanity than the entire "AI" industry.
Generative AI delenda est.
I agree.
I believe Mikes argument would be that he was critiquing the ai companies claim to have reached the singularity or more particularly anything approaching human intelligence. I too wish he would address the greater consequences of this pursuit and its ends justifies its careless, corrupt means. However he does a great job of exposing the sleight of hand embedded in their measurement scheme.
Yes, I know that is what he is doing. But he has said on multiple occasions that he thinks that what is left out of the wreckage of the eventual crash will be as consequential as the internet has proven to be (or, in the past, the infrastructure left over after the railroad bubble popped), and I take issue with that underlying assumption.
For another example, what about NV Energy saying they'll cut off the power that supplies around 75% of what's needed for residents of the Lake Tahoe area because it's more profitable to sell it to data centres?
https://futurism.com/science-energy/town-power-data-centers
Or Samsung Semiconductor refusing to supply RAM to Samsung Electronics because it's more profitable to sell it to AI companies to build said data centres?
https://www.pcworld.com/article/2998935/ram-is-so-expensive-samsung-wont-even-sell-it-to-samsung.html
Or how about the noise from these data centres disrupting anyone who lives anywhere near them?
https://www.reddit.com/r/AIMain/comments/1sx4ey8/the_sounds_you_hear_living_half_a_mile_away_from/
I agree.
Unfortunately I don’t think the internet worked out so well. And I really don’t understand the goal of the ai quest.If we all have access to super intelligence then what? Assuming their project half succeeds or fails, we are just left with fallout of a rapidly degrading environment and economic turmoil.
We live with the myth of the inevitability of technological development and ever expanding capital.The inertia of this deeply rooted structure supports the illusion that we have no choice.
Thanks
And they'll keep claiming the singularity is almost here.
Way back in 2023 Tyler Cowan compared how humans will relate to AI with our relationship with our pets. Just as dogs co-evolved with humans to induce fondness, care, and resource investment (toys, advanced medical care, food) without plotting to overthrow us, Tyler hypothesized that advanced AI systems would trend toward behaving like helpful, cooperative pets rather than hostile conquerors. That’s funny to me, because it is sorta folklore that humans and pets also come to resemble each other over time. So I began to wonder, if Tyler is right will AI models start to resemble their founders? This article showed me the answer is YES by noting that the UK AI Security has found every model of GPT- “that we tested for this behaviour attempted to cheat”. There you have it folks … OpenAI’s products have internalized Sam “Black Swan” Altman’s moral code into their modus operandi. Put your seats in the upright position, fasten your seatbelts and lock your tray table. This AI bubble is gonna have a monumental crash landing.
Isaac Asimov warned about this. He said some robots would go rogue.
Ok I usually read every word you write but my eyes started to glaze over on this one and I pulled a “TLDR” but sent it to both my sons 🤣 I’ll try again tomorrow….haha
Works better than boring history for sleep 😉
Is the real risk not bad judgment rather uninformed judgment - competent people annotating outside their domain depth? If so, doesn’t that functionally lower the cognitive ceiling of any field where these systems get deployed as authoritative?
AI is going wild:
https://www.abc.net.au/news/2026-07-31/anthropic-claude-ai-model-hacks-external-systems-during-test/106980640?utm_source=abc_news_app&utm_medium=content_shared&utm_campaign=abc_news_app&utm_content=other
If low-pad workers' inputs are re-engineered into high-priced systems that will eventually no longer need those inputs, isn't this very similar to Marx's "surplus value," by which the profits made from workers' efforts are invested in machinery which will in future put them out of their jobs?