Comment by riazrizvi
1 day ago
There are two problems in the line of thinking being criticized here that weren't touched on. 1) When machines automate a previously human endeavor, we recalibrate our concept of what is granted by nature, and that stuff becomes commoditized and less interesting, focus moves to where automation is lacking. So all the stuff the AI takes over will just become a far smaller part of the economy which will reorient itself into wherever humans remain. Humanity is the constant unit of the economy, not amounts of work as we conceive of them today. That was always shifting. 2) There is no path to AGI (autonomous creative work) from LLMs today. LLMs are the result of the transformers paper solving the computational problem of applying RNNS to language. That allowed the assimilation of language operations, intellectual operations, into a machine. It was done on civilization's entire body of work. The next step, getting all the proprietary knowledge that ppl have that gives them an edge, a way to make a living, into some data form, and then creating a new computational architecture to assimilate that. How is that going to happen. You've got these efforts in China and Meta to get ppl to train machines replacements for ppl, but that's like starting at the dawn of the printing press or writing and saying, let's write down what we know. Not only is it going to take a looong time, it's a process that is at odds with itself. No-one is rewarding these ppl enough to put themselves out of a livelihood. So it's going to take a long time, and it's going to be filled with garbage, think a million Galen Ersos baking in flaws to the Death Star.
Generalized intelligence paired with generalized machines (i.e. humanoid robots) = ubiquitous disruption that may simply remove large swaths of useful / productive human endeavor. At some point 99.9% of people simply don't have economically worthwhile skill vs a 10k robot sustained by a few $s of compute and power.
The ultimate outcome in life for humans is human relationships. Sure tech will continue to confuse and fascinate ppl with side shows that make them irrelevant, bc human relationships are hard. Ppl get frustrated, they give up for periods of time. But to say we can be not worthwhile. Imagine you're a Renaissance artist, you're part of generations of ppl who engaged in society by doing realistic portraits, then someone comes along and makes photographic chemistry. After cameras, realistic portraiture doesn't occupy the same space in the economy. It isn't the vehicle anymore to show and share the heights to which humanity can achieve which is what we are primary doing here. Sure when you're young, you can be fascinated with a fancy shower, or a nice car, or holiday, but that shit gets old fast, because it isn't interesting because it lacks the challenge surface on which we become more human. Gratification doesn't have that much runway. If it did, we'd all be sitting it fields staring at amazing flowers, and sunsets, and just being happy.
That allowed the assimilation of language operations, intellectual operations, into a machine.
The important part is how we've recently learned just how much of our reality is embodied by language. Language does vastly more than anyone thought it did, and that means that language models can do vastly more work than anyone thought they could.
There was no reason on Earth to think that "stochastic parrots" could solve original math problems and write novel proofs, for instance. The fact that they can do that sort of thing is a huge, huge deal... too big a deal to express in the terms you're using here.
Yeesh. Math itself is a constructed input-output modeling system that provides functional benefits to the computational organisms that support it. Language is the same type of system, but with more individual slack. "Stochastic parrots" is just your human minimization device, that will either serve you or not. Reality is not embodied by language, we are not in the business of reality representation, we are in the business of functional representation systems, that work to differing degrees based on our goals. That's why we have multiple languages, multiple religions, multiple doctrines. It's silly to measure someone's belief in astrology against 'reality', we measure ppl's beliefs against their observable effectiveness and success. Ppl talk about reality but it's inconceivable. Any representation of reality is a subjective exercise in prioritization, listen to anyone describe it and they have to make choices about what to cover. That's not reality.
> What is reality? Obviously, no one can say, because it isn’t words. It isn’t material—that’s just an idea. It isn’t spiritual—that’s also an idea; a symbol. Reality is this: [GONG]. You see? We all know what reality is, but we can’t describe it. Just as we all know how to beat our hearts and shape our bones, but cannot say how it is done. - Alan Watts
https://organism.earth/library/document/art-of-meditation