Iron Front

Through the Gates of Hell

  • Page 1
  • Archives

This is how Cyberpunk CEOs Talk

Fri 27 February 2026
By Ragnarokr

In Resist.

tags: techbrosslopobjective-realitystupid-statements

Screenshot from Shitter/X SAM ALTMAN: 'People talk about how much energy it takes to train an AI model ... But it also takes a lot of energy to train a human. It takes like 20 years of life and all of the food you eat during that time before you get smart.'

I've seen a lot of Techbro arrogance over the years, but this is a new low for Techbro arrogance. For a while, I just couldn't. Then I came to my senses, and tried to break down how much crap this actually is. Somewhere in the future that I hope we have, this will be in a book titled The Stupidest Things Ever Said, and Scam Altman won't be ruling over a substantial fraction of humanity in a bunker somewhere on the post-apocalyptic earth.

Fortunately, there has been massive retaliation across the internet. If you look at Altman's eyes while he says it, you realize he's perfectly comfortable with saying this with an utter lack of shame.

Analysis

YouTuber YongYeah has has an excellent teardown of most of this atrocity ("OpenAI CEO Argues Energy Is More Wasteful On Humans Than AI, Goes Very Poorly...") and some of the blowback.

Aakash Gupta (never heard of 'em before this) has an analysis of numbers (backup link). I don't agree with this analysis: it's still probably off by an order of magnitude since it doesn't include the embodied energy cost of food, clothing, etc for the humans OR the embodied energy cost and environmental damage of the hardware and infrastructure used to "train" the LLM. My guess would be that the latter is several orders of magnitude larger than the former:

  • Training models destroys components
  • Using the GPUs under heavy load (which would be the best way to get the most use out of them faster) causes the GPU devices to "melt down" (according to Scam Altman) while trying to rip off Studio Ghibli's style.
  • A given generation of GPUs isn't used to train one generation of a model and the next (the old ones are simply too slow for the increased cost of training "new" models.

Regardless, I'm not going to attempt to run the numbers, as that would be moral bankruptcy. A glorified spreadsheet and people are not morally equivalent.

Understanding This Through "LongTermism"

LLM company CEOs -- and many other techbros -- are best understood through their bizarre mindset, one component of which is a variety of of "Longtermism" or "Effective Altruism" which states:

  • There are humans alive now
  • There will be more humans in the future
  • At some arbitrary distant point in the future the number of humans alive now is (functionally) zero compared to the humans who will exist then
  • Therefore, no humans alive now matter

Maybe that interpretation is a little harsh. But this belief system is morally bankrupt: if you accept that:

  • two people matter more than one in all abstract senses
  • there will always be more humans in the future than now

... then the natural consequence is that no humans currently alive will never matter. 1

More links on this variety of Longtermism:

  • Doomsday Bunkers, Mars, and 'The Mindset'
  • A dangerous calculation
  • The 'dangerous' promise of a techno-utopian future'
  • TESCREAL if you read nothing else here, read this

What Sam Altman is Actually Saying

Principle: when someone shows you who they are, believe them.

Some of this comes from a David Fairchild tweet (backup link)

Humans only matter where they can generate shareholder value for me and my elite club. Until you're educated (with the right kind of education) you're just a useless eater. If I can consume gigawatts of electricity to replace you with a synthetic intelligence, then it won't matter if you don't have a liveable biosphere. Meat humans are just one more stepping stone for me.

  1. This bears a remarkable similarity to conservative arguments against foreign aid, and their refusal to also help needy people 'here' ↩︎

Categories

  • Observations
  • Resist

Blogroll

  • Cory Doctorow
  • Sciop
  • RationalWiki
  • IFixIt Canada
  • The Beaverton
  • Venn Diagram of Irrational Nonsense
  • How to Survive the Broligarchy
  • Pepper & Carrot
  • Bellingcat
  • Made with Pelican

Social

  • atom feed
  • rss feed
  • Contact Email (rotated frequently)
No rest, no mercy, no matter what. Still under construction.