Skip to content
DevMeme
6013 of 7590
AI ML Post #6582 · source on Telegram

When AI progress graphs forget there’s a second hyperbola branch

Description

Meme shows a Cartesian grid with a red hyperbolic curve plotted against time on the x-axis (labels: 2020, 2025, 2030, 2035) and unlabeled y-axis gridlines ranging roughly from - 5 to +∞. Above the plot, black text reads: “everyone seems to be saying AI progress is gonna be hyperbolic.. but if so wtf does this part mean??”. The curve hugs the x-axis until ~2028, then shoots vertically upward toward infinity just before the 2030 tick; the complementary branch dives from - ∞ just before 2030 and rises back toward zero by 2035. A thick yellow oval highlights this lower (negative) branch, implying confusion about what a hyperbola’s ‘other side’ forecasts for AI. The humor riffs on hype-curve extrapolations, mathematical literacy, and the tendency to assume unbounded exponential-style growth while ignoring inconvenient asymptotes and inverse branches

Comments

14
Anonymous ★ Top Pick That circled region is the CFO’s budget line after you scale model parameters past 10¹² - turns out the singularity has a balance-sheet antipode
  1. Anonymous ★ Top Pick

    That circled region is the CFO’s budget line after you scale model parameters past 10¹² - turns out the singularity has a balance-sheet antipode

  2. Anonymous

    Ah yes, the classic AI hyperbolic growth curve that becomes so advanced it invents time travel, goes back to 2030, and decides to tank its own progress after realizing it left the GPU cluster running and the AWS bill exceeded the GDP of several nations

  3. Anonymous

    Ah yes, the classic mistake of extrapolating hyperbolic growth without considering the full domain - turns out when you model AI progress with y = 1/(x-2028), nobody mentions the asymptote works both ways. That downward curve after 2030? That's either the heat death of GPU clusters, or the moment we realize we've been overfitting on benchmarks this whole time. Perhaps the real singularity was the mathematical functions we misunderstood along the way

  4. Anonymous

    Post-spike drop: when trillion-param models hit the Bekenstein bound and start entropy-encoding their own obsolescence

  5. Anonymous

    Everyone says AI growth is hyperbolic; great - are our OKRs on the branch before the 2030 asymptote, or the one after it where progress < 0 and the budget divides by zero?

  6. Anonymous

    If AI progress is truly hyperbolic, we race toward 1/(2030−t) infinity - then discover budgets and regulation are discontinuous functions, landing on the other branch; in other words, welcome to production

  7. @kennyotsu 1y

    Back to sticks and stones

    1. @patsany_horosh_mne_v_dm_pisat 1y

      Exactly

  8. @seyfer 1y

    World population

  9. @Art3m_1502 1y

    The machines rose from ashes of the nuclear fire...

  10. @HarukiNikaidou 1y

    world war III with huge amount of nuclear weapon use

  11. @azizhakberdiev 1y

    integer overflow

  12. @mira_the_cat 1y

    that's because no one knows what will be after, maybe it's like negative temperature that's actually hotter than any positive temperature 😁

  13. @ercolebellucci 1y

    Asynshit

Use J and K for navigation