474 karma · joined September 1, 2020
Edit: Ah: there's a quiet hamburger menubar. The page really should have forward/next/toc links.
Humans employing this kind of thinking and desire to control other people is the genuine danger here.
It feels like we have collectively goal-post shifted the definition of AGI to be closer to that of ASI.
One of those "what do you call a doctor who graduated at the bottom of their class" type things. I think despite their genuine deficits (and there are many), frontier LLMs have basically cleared the minimum AGI bar.
Though I realize a lot of people don't agree with this take :)
You must have a different definition of most of the words you used than I do.
I'd wager 99.99% of humanity would not want to start a nuclear war by vaporizing millions of people, animals, etc. with a nuke. They'd back away slowly and figure out how to ensure it was safely disposed of so they didn't need to be near it or worry about it.
A very small number of people would like to have a nuke with which to harm other people.
Some of those people even probably got that way because of actions taken by people saying things like "if we don't do it, somebody worse will" and "we are the only ones who can be trusted to control this power"
Without moving goal posts: @27183: the word intelligence is defined as "noun. The ability to acquire, understand, and use knowledge."
LLMs can absolutely do this via in-context learning. They can even persist this as mementos on disk and re-attend to it later if its been purged from their context window.
You might try to make some comment like "they're not learning in the way humans do, they're not changing their weights, and they predict token by token!" Who cares? That's semantics - LLM attention is able to accumulate contributions of tokens and weights mixed together head by head and layer by layer with non-linear activations which functionally quack like intelligence.
They may be far different from us, but they're functionally intelligent (at least in several subsets of intelligence. And frankly, in some cases, they're more functionally capable than many humans).
Birds, planes, helicopters, rockets and gliders don't achieve flight in the same way, or have remotely the same capabilities in the air - but all are nonetheless forms of flight.
When an LLM makes a cognitive mistake, humans jeer "dumb stochastic parrot!" - when a human does the same, many are blind to it. Like a weird auto-Gell-Mann Amnesia - relative to reflecting on our own perceived cognitive "strengths" versus others weaknesses.
I think maybe you mean superintelligence, which is a more fair critique.
Lest you think I'm being hyperbolic: https://ai-2040.com/supplements/covert-ai-projects
This is arsonists selling fire insurance.
I saved money and bought a 2400 baud modem from a classified ad for $25 (had to drive a couple of towns over to pick it up) in something like 1995. IIRC, it was well over a hundred dollars for a 14.4 modem at whatever time that was. My whole 8086 PS/2 computer was salvaged parts, paid for by my own savings from my underage labor.
https://askubuntu.com/questions/1498611/ubuntu-dhcp-client-u... (linked because depending on version, there are several different ways to make this change..)
"However, these inference optimizations, which rival Anthropic refers to as “compute multipliers,” are a big focus for all the labs. Anthropic CEO Dario Amodei has been publicly talking about the concept since at least mid-2023, when he said on a podcast that the company limits “the number of people who are aware of a given compute multiplier” because it could give other AI labs a leg up if they were to be able to replicate them. (Compute multipliers can also refer to efficiency optimizations in the model-training phase.)"
Yes, on a world with finite resources where your industry is singlehandedly siphoning ALL THE RESOURCES - hoard general efficiency optimizations and treat them as trade secrets - winning is all that matters, normal people and other species and the planet be damned.
Everything I hear about Dario these days makes me like him less and less. He sure did seem to speed run the 'tech leader with scruples' to 'tech villain' path! I guess all the cycles are compressing as we approach the singularity..
The problem is that politicians were corrupted by power.
We need to stop building surveilance panopticons!
"it is even conceivable that Norris’s pixels could react to a captured image and, without going through a computer, produce corresponding light patterns."
Great, also they invented a digital mirror (and digital fun house mirror).