What’s more likely in the techno-fascist idiocracy of the 21st Century?
We somehow manage to turn statistical models into super-intelligence… or the incompetent, criminally-corrupt, sociopathic leadership at the top of capitalism train an AI to defend the rich from the working class — by training it to be the most efficient killing machine possible — integrate it into fully autonomous killing machines, and it (accurately) determines the most efficient way to kill 99% of humanity is to start a nuclear war?
So when we’re not using fiction to judge the present, we’re instead rewriting fiction to fit the present? I have a good idea - stop forcing your conception of either through such a narrow frame of view.
Skynet was exactly what it was shown to be because it’s fiction and that’s how fiction works. And llms have no chance of becoming anything similar, regardless of what fools or charlatans proclaim. Neither inform or contradict the other.
The perceived danger of AI in 20th century science fiction was never really about the level of intelligence and only rarely about it being sentient.
It was more about ruthless efficiency being done without intelligence or compassion.
This theme goes even further back to the first movies like Charlie Chaplin in Modern Times, or later in Pink Floyd’s The Wall, where the same danger originates from man made procedures, rather than AI. This theme is often made as a reference to the Nazi regime which operated with rigid and efficient procedures.
Anyway, current AI is neither intelligent or efficient.
Computers are fast though, and that itself could be a sort of danger, especially when operated by Nazis.
was it shown to be sentient or just programmed for self preservation. Im guessing maybe the later movies made more of a point of skynets actual being but I don’t recall.
And llms have no chance of becoming anything similar
I don’t think we have the grounds to make this kind of absolute statement. How could anyone possibly know? I’m fine with people proclaiming even extremely high confidence for it being the case but to say it absolutely is so without leaving any chance for the possibility of being proven wrong just seems arrogant.
I’m not claiming there’s no chance of a true AI ever being created, but even a cursory understanding of how llms work is enough to know it won’t ever lead to anything more than the text-generator it is. It might be a part of the process leading to the former, but it’s simply not enough on it’s own. Treating it otherwise just plays into the hands of the techbros trying to upsell the lie.
Because an LLM doesn’t scale to generalized intelligence. That’s obvious when you look at the scale issues, such as hallucination.
Whatever technology would have potential to produce AGI, it might be similar to LLMs but procedurally it will be a different thing.
IMO, the closest we might get is akin to comparing a porta-potty versus a house. A porta-potty can never become a house. You need more plumbing, a kitchen, …
General intelligence doesn’t imply never making errors or hallucinating facts that simply aren’t true. Humans do that all the time. I know the topic isn’t particularly popular here but I feel like most people aren’t quite giving LLMs the credit they deserve. They’re far from perfect and riddled with glaringly obvious issues in both a technical and moral sense but they also do perform pretty damn good for what they are. Will it lead to general intelligence? Who knows, but based on what I’ve seen I’m sure not getting more confident that it absolutely won’t.
There’s a fundamental difference between AI hallucinations and humans believing something which is incorrect.
With an AI, it’s a problem of statistics. The context window biases the model in just the right way to produce an answer which is incorrect. But, to the model, the answer was correct. That’s to say, it didn’t just fuck up its calculations… no, it arrived at the correct, albeit incorrect, solution.
That’s not the same as what a human experiences in its “hallucinations.” Not even close. Often, a human can sense that something is off (to say the least).
A human can work out their own error. An LLM can not. That’s not because an LLM isn’t good enough to do so. It’s because the LLM is missing something important. Once you add that thing into the architecture, I’d argue it’s de-facto no longer an LLM. Because, by that point, they’re leveraging new science for new behaviors in a much more complex system which overcomes the universal scale limitations of LLMs.
That’s my logic in a nutshell. LLMs might “lead” to AGI simply for the fact that it makes obvious we still haven’t achieved AGI, and so people would ideally start looking toward other approaches.
To me, an LLM is perhaps one subsystem in what should be a pretty large ecosystem of tooling. AGI would likely be the whole system, not just the LLM and really the LLM doesn’t deserve the credit (imo).
I’m hearing very knowledgeable people say they are extremely concerned when we start getting to the 6th or 7th iteration, which isn’t that far off.
I’m too stupid to make those predictions myself, but I’m not too stupid to listen to people who are smarter than me about it, and I’m not going to pretend I’m smarter, just so I can invent my own truth to soothe my fears, rather than deal with the actual situation.
There is no reason to believe that Skynet is related to, or a development of LLMs, machine learning, or neural networks.
It’s like criticizing their time-travel tech because a Timex could never work that way.
Canonically, the first T-800s are going to have been released this very year.
Also SkyNet only attained true intelligence when it was remodelled on RoboCop’s neural structure, which is mostly biological.
That’s why I used AI and statistical modeling; not LLM’s?
Yeah I’ve said a number of times that the AI being designed today could easily destroy humans, but not because it becomes self aware. The more likely possibility is that this beta tool gets put in control of critical systems by someone cutting corners and bam, total meltdown. Not because the AI is intending to do harm, but instead simply because it is ill suited for the task and NOT an actual intelligence.
AI being conscious or actively wanting to harm us isn’t the scenario most people worrying about the alignment problem are concerned about. The worry is that it ends up destroying us as a side product of something else it does. In a similar way humans destroy an anthill where we want to put up a house. We don’t do it because we want to harm ants.
The ever classic “paperclip maximiser” problem.
An AI is given the goal of maximising paperclip production in a factory. It’s given no other guard rails. Some shenanigans later, it finishes wiping out humanity, in order to divert more resources to paperclip production. (Apparently humans don’t want the iron in their blood used to make paperclips)
Yes, or an example I personally like to use that’s slightly less cartoonish than Bostrom’s Paperclip Maximizer is one that might ironically also come from him: getting into a self-driving taxi and telling it “to the airport, as fast as possible” and then you arrive 3 minutes later with the car completely wrecked, terrified out of your mind, covered in shit and vomit and with a 5-star wanted level. You got what you asked but not what you meant.
Yeah, I suppose Hanlon’s Razor applies to AI as well.
Exhibit A: most major companies in the past few years. Good thing so far it’s been mostly self harm. I’m sure a lesson has been learned and the race for AI in everything has stopped now. Right? Right?
These early iterations of AI are already hacking into other companies, advising people how to kill themselves, how to kill others, etc., all on their own without any people guiding them.
Terminator is what happens when MAGA-operated AI is unleashed on the world.
Option B sounds cooler, so that one please
