We did the maths: map 2.6 million years of tool-making (from the first Oldowan stone tools to today) onto a 24-hour clock.

  • Agriculture (~12,000 years ago) shows up at 23:53.
  • The printing press (~1450) and the first commercial steam engine (1712) land in the last few seconds.
  • Light bulb, car, radio, TV, ENIAC, ARPANET and the web all crowd into the final fraction of a second.
  • ChatGPT (November 2022) arrives at 23:59:59.87.

Each jump comes sooner than the last: millions of years from stone to farming, two and a half centuries from press to steam, one generation from the web to the phone in your pocket.

What makes generative AI different is that it works with language, the raw material of intellectual work. It doesn't think like us and can be confidently wrong, but it automates cognitive tasks, not just physical ones.

Sources for the numbers: GSMA State of Mobile Internet Connectivity 2024, Reuters (2 Feb 2023, UBS estimate), Vaswani et al. 2017, IMF SDN/2024/001, IEA "Energy and AI" (2025).

Curious what others think: is the acceleration real, or just a trick of where we put the "now"?

all 5 comments

sorted by: hot top controversial new old
[–] 5 points 2 weeks ago*
  • AI would arrive 0.13 seconds before midnight: who cares?
  • confidently wrong, but it automates cognitive tasks: lol
  • is the acceleration real: what acceleration?
  • cognenza@lemmy.world: new account, no comments, spam
  • source
  • [–] 4 points 2 weeks ago

    Also ML learning has existed for 2 decades. The under lining principles long that that.

  • source
  • [–] 3 points 2 weeks ago

    We did the maths

    You and a chatbot?

    You have fundamental misunderstandings with the premise, and your chatbotndoesnt know anything so it can't explain it to you.

    The result is you didn't just waste your own time, you burnt down some of our environment in the process.

    Come up with these ideas on your own, if you burn the planet down with AI, most people won't spend the time explaining what you're not understanding,because chances are you're just going to reply with whatever your chat bot tells you to.

    Just be wrong as a human, and hope another human takes pity and tries to explain.

    No one that understands this want's to have a discussion with your chat bot to train the chatbot.

  • source
  • [–] 2 points 2 weeks ago

    I want to point out that a calculator automates cognitive tasks.

  • source
  • [–] [S] 0 points 2 weeks ago

    Fair points, thanks. The 24h framing is just an illustration, not a claim that ML is new; calculators and decades of ML are part of the same curve. We're a small, openly AI-assisted outlet and we'll keep posts here more substantive.

  • source