Opinion
AI, the Theory: As it grows, it grows exponentially, until it eclipses the human brain.
AI, the Practice: As it grows, it grows asymptotically, until it grinds to a halt under the weight of its innate limitations.
- - -
People on Bluesky had a hoot upon learning OpenAI's GPT-5 [which I've, for a hoot, nicknamed Generic Plagiarism Technology]
- Is behind schedule
- Suffers high computing costs
- Has access to limited high-quality training data
(according to the Wall Street Journal [source]).
- - -
seed-corn-thoughts: Well, well, well, looks like we're probably at the top of the exponential growth curve for AI models, which, in turn, means that they'll be, at best, what I said they'd be: labor-saving tools, but nothing that can legitimately replace creatives.
the-pizza-dude: I run a business (when I can be bothered). Number of times I call my accountant? Every couple of weeks. Number of times I've used [Microsoft] Copilot? Still trying to delete/disable it.
Audrey Truschke: Almost like AI, in addition to being a massive copyright violation, misleading, and vague, is making us dumber. Leading to “limited high-quality training data.” I'm telling y'all: The best thing we can all do in the humanities, and as human beings, is to stay far away from this dumpster fire.
Jeffrey P. Bigham: The “easy to get data” has been gotten. This is why things like world models and such are actually interesting. If you've run out of nicely prepackaged human data, you gotta start generating new data by interacting with stuff in sufficiently complex simulations or the real world.
doormat9: Love waiting around to see if this is all going to go the way of 3DTV, or end civilization.
Benjamin Riley: There's a lot to chew on in this story, but it's inescapably amusing to me that the desire for “more data” to feed into LLMs [large language models] is fueling efforts to hire humans to produce knowledge. The thousand-monkey/thousand-typewriter theory of software development, only we are the monkeys.
András Forgács W: Well. Maybe it's time to get out of the AI thingy before you lose all your investment.
bazlyons: They should ask AI how to fix it if it's so smart!
Gramsci ZA: So, it appears they've nailed the artificial part, now a few more billion to sort the intelligence bit.
crankgrimes: Yet another exponential curve turns out to be sigmoid [S-shaped].
Stephen Rowley: The LLMs are plateauing, and are going to be stuck at “incredibly expensive and largely useless.” I cannot wait for this bubble to burst.
vietdongsoldier: Been a whole series of these stories now, and it looks like the investor class is finally getting nervous about the chances of seeing meaningful returns on their AI investments.
Nicolai B. Hansen: Who could have thought that BIGGER COMPUTER AND MORE DATA wasn’t the solution, nobody ever said that [taps Searle, Dreyfus etc sign].
Tom Hearden: Meanwhile somehow the “valuation” magically rose from $87 billion to $157 billion over a 9-month span this year.
Bennett Tomlin: I think one $10b fundraising round will fix this.
[Source: https://www.techmeme.com/241220/p31#a241220p31]
Comments