“But it also takes a lot of energy to train a human,” Altman said. “It takes like 20 years of life and all of the food you eat during that time before you get smart. And not only that, it took the very widespread evolution of the 100 billion people that have ever lived and learned not to get eaten by predators and learned how to figure out science and whatever, to produce you.”
So in his view, the fair comparison is, “If you ask ChatGPT a question, how much energy does it take once its model is trained to answer that question versus a human? And probably, AI has already caught up on an energy efficiency basis, measured that way.”
Chatbots aren’t the endpoint tho, AGI and ASI are. Imagine a future where we could disseminate custom AIs to teach kids exactly in their unique contexts. Education could be throttled and specialized according to everyone’s aptitudes.
And if the people are able to decide what future ASIs work on we could focus on massive healthcare, leading to inevitable healthspan extension. Then the rate at which we have to replace our population (and the associated spending of 20 years reteaching intelligent citizens) would be reduced as a consequence.
AI is just a tool. Tools are never rolled back, at most they’re only regulated. Why not make the best of our future with this powerful tool? Just because billionaires are getting the headlines, most of the progress is being done in academia. Maybe AI will even help facilitate reduced wealth inequality.
That assumes the AGI/ASI doesn’t have gates keeping it from people. Sorry you only get the basic model b/c you can’t afford it something better. An open, free, equally accessible AI for everyone no matter what is just not how capitalism works.
You’re assuming the alternative that billionaires, despite being a tiny fraction of the population, will be in control of such gatekeeping.
Your argument against an openly available AI precludes the existence of things like the FOSS community. Smart people who oppose capitalist power structures certainly exist.
I want kids to have social skills, so I will not imagine this.
In what way does AI detract from social skill development?
You learn social skills from people.
In what way does AI prevent people from socializing with one another?
You’re asking how an AI that tutors you might displace the need for people who tutor you? Hm.
You know that people used to know their tailor? You could bring your son in, get him a suit for his birthday. “Hey, this is my boy, William. William, this is Richard, I’ve been coming to him for years. He does great work. How’s the wife, Richard?” I have never spoken to the Amazon website like this. I don’t think it can hear me.
Steelmanning is what wins arguments. For example if I were to say that your argument amounts to little more than lazy contrariness some percentage of us Lemmings would see that as uncharitable, regardless of whether they agreed with my position. I’m not suggesting updoots are important, just that discourse is.
That being said while education and socialization aren’t inherently dependent on one another, certain subjects like debate, civics, and ethics should likely be taught in group settings (as well as more often). PhysEd as well.
But if harder sciences and math have the potential to be taught outside the sometimes stressful social hierarchies of traditional schools, it’s worth at least exploring.
PS: Regarding your username, have you seen the excellent Philip Seymour Hoffman movie “Love Liza”?
Actually, no, it’s not. Steelmanning is a tactic.
I mean, as long as people believe in the truth, it’s a tactic I’d hope they use.
The main reason that I’m not engaging with the paragraph-by-paragraph is that I just don’t feel like taking 6 months to explain a systemic view of society. It is obvious to me that you view an AI education as no different than opening a textbook.
I don’t know how you live, so let me tell you something. If you wanted to, you could: work from home (to be fair, I love doing that), get all your trinkets and toilet paper from Amazon, spend your off-time watching vtubers pretend they’re uncomfortable with the word ‘penis’, order all of your food through UberEats, talk to people, if you do, exclusively through Discord and Reddit, ignore all phonecalls and have AI write your texts back instead, skip bathing entirely because what purpose does that even serve at this point, and spend the rest of your time being intellectually stimulated by gacha-game roulettes and call of duty lootboxes.
For a lot of people, school is the one time when they can’t do this. They are forcibly dragged by their heels over gravel and concrete into a community with other people. This is a place they can look up and see other faces. They can stand in places where their obscene body odour is shameful enough that they may be bullied into bathing again. They could, perhaps, see another person struggling and offer to tutor them.
Yes, technically, you could educate from an AI, and separately, socialize with people elsewhere. Technically, and I’m not saying these are strictly equivalent, you could segregate your schools by color, but still offer your jobs equally to any applicant, whoever they may be.
The question I ask is, “but would they?”
The social ramifications of teaching people to be dependent on this technology, and this includes their social skills, their sense of community membership, where they feel like they get their friends from (the sycophantic chatgpt is much, much better at affirming your bad habits than any person will be), are so dangerous that I don’t really want this technology anywhere. Certainly not in a classroom before anyone has even learned to be self-actualized.
I would rather imagine a world where teachers are paid well. Where more faculty can be hired. Where classroom sizes are systemically allowed to be smaller than they are. Where no-child-left-behind laws, which are destructive, are broken and shattered to pieces. Where students form study groups and support each other, something they should be doing their entire lives, instead of asking a T-1000 that pretends it can giggle.
This is a really basic one, probably not viable for the modern age: every time you need to ask a person a question is an opportunity to make a friend. If you are asking your questions of an AI, where are you making your friends? I’m not implying an answer, this is an open-ended question.
PS: No, but this summary is really fucking funny:
Take off your rose-colored glasses. The billionaires control the implementation, and they don’t care about universal healthcare or education, and certainly not reducing wealth inequality.
Also, LLMs will never evolve into AGI, because they’re fundamentally different processes. The human brain has many parts, and the Broca’s area is a very small part that cannot generalize its functions to perform the tasks of all the other areas of the brain.
And ASI is pure scifi. Circuits will never be sentient, the best they can ever do is mimic sentiency.
The most common way people give up their power is by thinking they don’t have any. The corrupting nature of power is known by the majority. Billionaires just ignore it because capitalism rewards executives who exhibit psychopathic symptoms.
I don’t believe anyone’s arguing LLMs will evolve into AGI.
Thats a conclusion of substrate dependency. What’s so special about the matter (not the process) that facilitates human cognition that makes it impossible to happen with other materials?
such a simp