I have been hearing people say "[latest model] is AGI already, organizations just aren't leveraging it yet" all the way since GPT 4o. And people will probably continue making this claim every time a new frontier model comes out.
I have too, and I can't help but notice that the people saying this also happen to be quite eager to sell me the secret sauce for successfully leveraging it.
The thing they never seem to want to talk about is how the world already has several billion instances of non-artificial general intelligence. And, while they can collectively do quite a lot of different tasks, each one is only competent to do a vanishingly small percentage of all possible tasks. And that it usually takes quite a lot of work to get them to the point of being able to independently perform most of the interesting tasks. Because if we did talk about that then we'd have to start answering questions about why we expect artificial intelligence not to also be like that.
It's AGI. It's just stupid sometimes and genius sometimes as well.
We like to emphasize the stupid part and use that as proof agi hasn't arrived because it's comforting. Don't want to look at it while it eats us alive.
This is why I kind of hate the term though. We generally agree that humans are intelligent even though they're also pretty damn stupid.
For example, one of the stupid things we do is fail to realize that "intelligent" is a word with multiple senses. Which then makes a muddle of conversations like this because we all - myself included - tend to switch back and forth between senses of the term without even realizing we're doing it. So we have sense A of "intelligent", which means something like "capable of abstract reasoning". And we have sense B of "intelligent", which means something more like "the opposite of stupid". These two meanings don't actually have much to do with each other. But a lot of the hype around LLMs and the possibility of them being a form of AGI tends to hinge on tacitly assuming that an artificial system that can do abstract reasoning tasks won't do stupid things. Simply because sense A has one characteristic, sense B has the other, and we happen to use the same word to describe both.
I think that, yes, it is rather unfair to hold AIs to a higher standard than humans in order to reach the "AGI" label, even though we consider humans to be "general intelligences".
But the definition I had in mind was the one in Wikipedia, "matches or surpasses human capabilities across virtually all cognitive tasks." Well, if it's going to do that, it can't be stupider than capable humans nearly as often as it is.
Heck, by that definition it needs to be able to count to 100 and know when it’s appropriate to use a syllogism.
It’s things like that that leave me convinced it’s not really AGI. The symbolic reasoning element doesn’t really seem to be there. And I’m not prepared to call something general intelligence if it can’t do that.
Also, maybe even more to the point, it can’t even learn. No, pre- and post-training are not the same thing.
Sashiko, the LLM patch reviewer that now comments on kernel mailing lists, merged a fix of mine: https://github.com/sashiko-dev/sashiko/pull/556 (a patch that contained one of its prompt placeholders as text had it replaced).
ntfs: writing to a compressed file on a volume whose free space is fragmented silently overwrote clusters belonging to other files. write() reported success; the damage only showed after a remount.
ntfs: reads of a damaged file could return zeros for data that is on disk, and writes to it could be dropped.
Sashiko: on its Claude, Vertex and Gemini backends, a connection that drops mid-response is treated as a permanent error rather than retried.
The rest, mostly more ntfs, are still in review or on my desk. Not all of these came from that first night; the ext4 and ocfs2 ones are older.
I also have a lot of bcachefs contributions, but that's because I'm actively using that filesystem on my desktop; instead of a pure fuzzing run.
That explanation doesn't make sense. If the point is anonymity, why not just AI generate the students entirely? They AI-generated the girl on the right in her entirety, so capability clearly isn't the problem. And the guy in the middle is still very much recognizable.
I don't understand what you mean? For me, it appears all 3 students have been changed. The 3rd student is changed into a black female. I am not doubting they want to inject diversity (although i didnt say it above) my contention is they likely view the original as a blank background, which they inserted 3 "new" anonymized students into. Its not that they turned 1 person "black", its that they changed ALL 3 PEOPLE entirely
Given how little the two students on the left changed, I highly doubt the AI made a creative decision to completely randomize only the student on the right. I think it's more likely that they did it in two steps: tweak the faces, and replace the student on the right.
But what if (as the commenter you're responding to stated) someone believes that AI will have a positive effect on humanity? Is it morally wrong for them to disagree with your view?
I personally think taking part in an industry that openly and casually argues that the insane losses being racked up are fully affordable as long as people are permanently taken out of the workforce is immoral, yes. Pan-industrial job loss is a harm.
Leaning on "I did what I thought was right for humanity" while not addressing the harms to people along the way is not moral.
Personally, I don't know if AI will be a net-positive or net-negative for mankind, but if it's negative it's overwhelmingly likely that it will be due to environmental and/or energy usage concerns. That is, we ought to limit the power consumption and force companies to work more efficiently and sustainably, rather than limiting the research itself.
Arguments as old as time that a new form of automation will put people out of jobs are just not compelling to me, I'm sorry. I also don't believe it's likely that LLMs will ever be ASI, nor ever evil and sentient.
> Arguments as old as time that a new form of automation will put people out of jobs are just not compelling to me, I'm sorry. I also don't believe it's likely that LLMs will ever be ASI, nor ever evil and sentient.
I didn't make either of these arguments.
I was literally talking about what "bad" means, contextually, in the context of a job, responding to the idea that most people just don't see "good/bad" issues in their jobs. When almost all of us do, in fact, see harms we should reduce. Just often smaller harms.
One of the problems in the tech industry is that people always say "well sure, but…" and then try to expand the scenario to an amoral Thanos-level so they can wriggle off what might actually in many cases be much more personal, contextually obvious hooks.
"It happens, I shouldn't worry about my part in it" is not a moral framework.
They're not challenging that at all. The time limit makes it a game rather than an experiment. "You can't tell it's AI if you don't have enough time to examine the details" is not exactly a compelling claim...
When I was doing arbitrage and bonus farming, the process often involved wagering both sides of a bet on different sportsbooks. Result being that my DraftKings account was eventually down thousands of dollars over its lifetime.
So of course... they emailed me offering access to a VIP bonus program! To get in I just needed to keep betting as much as possible. Such nice people.
reply