Hacker News Reader

> And while LLMs are certainly an exciting new technology, it's not at all clear that they're really more than a glorified autocorrect.

Are we sure things like biology, or heck, even the universe as a whole and its parts, aren't "glorified x thing"? Can't we apply this argument to just about anything?

I feel like comparing it to biological systems glances over like half the points the parent makes. Also of course it's a sophisticated autocorrect, there's no system for agency, which is what we also desire from a proper AGI.

I wanted to put the focus on the overused "glorified x thing" sentence, which to me seems to be applicable to just about anything. I didn't want to liken/compare AI to biological systems per se.

Free will is an illusion. no one has agency. we're all just following our baked in incentives.

I'm able to take initiative and act on my own internal thought processes, though. I'm not limited by whether someone prompts me to do something.

You and me think we are but we can't be sure, and many before us have raised a doubt.

As for the prompt(s), to use such a limiting term, they could as well come from a self-reinforcing loop that starts when we're born and is influenced by external stimuli.

LLMs as part of a bigger system that keeps prompting itself, perhaps like our internal conscious thought processes, sounds a lot more like something that might be headed towards AGI. But LLMs on their own aren't it.

I have a theory (with no data to back it up; would be curious to get people's thoughts) that people with a religious or spiritual world-view, who believe that there is such thing as a soul, and that the mind is more than just a collection of neurons in the brain, are much less inclined to think that "AI" will ever reach a sort of singularity or true "human-like" intelligence. And likewise, those who are more atheist/agnostic, or inclined to believe that human consciousness is nothing more than the patterns of neurons firing in response to various stimuli, are more convinced that a human-like machine/programmed intelligence is not only possible, but inevitable given enough resources.

I could be wildly off base, but seeing many of the (often heated) arguments made about what AI is or isn't or should or could be, it makes me wonder.

As it happens, I am indeed Christian. But I see the soul as the software that runs on the hardware of our brain (although those aren't as neatly separated in our brain as they are in computers), and I suspect that it should be possible to simulate it in theory. I just think we're nowhere near that. We still don't agree on what the many aspects of intelligence are and how they work together to form the human mind. And then there's consciousness; we have no clue what it is. Maybe it's emergent? Maybe it's illusion? Or is it something real? I don't think we'll be able to create a truly human-like intelligence until we figure that one out.

Although we're certainly making a lot of progress on other aspects of intelligence.

And then there's all the talk of a singularity in innovation or progress that to me betrays a lack of understanding of what the word singularity means, and a lack of understanding of the limits of knowledge and progress.

Of course, when you don't define your X, you can use that phrase for anything. That's trivial logic.

I don't define X because it's highly variable.

It seems to me that on one extreme there are people easily anthropomorphising advanced computing and on the other extreme there are people trivializing it with sentences like "glorified x thing". This time around it's "glorified autocorrect" and its derivations. It's always something that glorifies another artificial thing, and I suspect that if and when we will have recreated the human brain, or heck, another human, it will still be a "glorified x thing".

As 0x0203 said, maybe it is to be ascribed to the religious substrate that takes offence at anything that arrogantly tries to resemble the living creatures made by God, or God himself.

It wouldn’t. It would be a logistic curve. Pretty much everything people call exponential should actually be logistic

That just agrees with the post. Nothing becomes vertical or infinite, at best exponential for a limited time, i.e. logistic.

I’m aware that it agrees with the post

Its infinite from the perspective of our side of the curve.

It's another application of advanced technology appearing to be magic, but imagine transitioning into it in a matter of hours, then with that advanced tech transitioning further into damn-near godhood within minutes.

Then imagine what happens in the second after that.

It may be operating within the boundaries of physics, but they would be physical rules well beyond our understanding and may even be infinite by our own limited definition of physics.

That's the curve.

I think it's completely ridiculous how you people just casually imply that a current computer can somehow achieve "damn-near godhood".

Hardware is a bottleneck. Better hardware can't be designed, manufactured and installed in minutes.

You can't just infinitely improve software sitting on the same hardware. At some point the software is as efficient as it can possibly be. And realistically a complex piece of software won't be close to optimal.

You people love dreaming up possibilities but you never stop to actually think about the limitations. You just assume everything will work out magically and satisfy your fantasy.

I'm glad to see people like you that understand that computers are ultimately machines that we put electric current (limited by our infrastructure) in to, and even with optimal efficiency (which we are no where close to), some math comes out. This year we all learned that a lot of what we thought was human thought can be expressed as math. Cool. But we can't scale that any higher than the physical input to the computer, and the physical heat dissipation that we can handle either. A 100% occupancy computing device is effectively a radiator.

So singularity people think 2 fallacies:

1. That it is possible to write an exponential function that somehow scales to asymptotic

2. That it is possible to write software that is so energy efficient that its output somehow exceeds the power draw

1 can be redefined to say "effectively" vertical. Ok. Cool motive, not a singularity.

2 can be handled by pumping more and more power in to the machine and building new nuclear plants to do so. This is what Microsoft is doing today.

Science inherently concerns itself with understanding and defining limits. However, when it comes to intelligence, there appears to be a significant gap in our grasp of these boundaries. Reflecting on my own thoughts from 35 years ago, I recall envisioning the singularity at the age of 12 or 13. Looking back, I realize that many of my conceptions from that age were either incorrect or grossly exaggerated. This realization leads me to speculate that the notion of the singularity might fall into a similar category, although concrete evidence to support this is currently lacking. While I find myself partially aligning with your points regarding magical thinking in this context, I am curious about how one might construct a robust, scientifically grounded argument in this domain.

> Hardware is a bottleneck. Better hardware can't be designed, manufactured and installed in minutes.

Yes you can. The Cloud.

Even the cloud has its limits, though. It's not infinite, and it does require money.

True, but it does make hardware much more available and not a bottleneck anymore. At least not in the same way. At some point, if it hasn't been reached already, the power available with existing hardware will represent more compute than is needed to hold AI. At that point AI could start creating its own hardware production facilities as well.

Like the Rapture and other concepts like that. How fascinating. Within a split second, in the time it takes someone to blink and hear a trumpet's blast, everything could change. Amazing! I can't wait!

As both a Christian and an AI-educated software engineer who subscribes to neither the rapture nor the singularity, I really like this comparison. They both feel like a kind of escapist thinking that in the near future, this magical event will happen that will make everything else irrelevant.

The Singularity people and the Rapture people are both correct. Can already see the event starting to take shape, just gotta know where to look for all the signs. Everything is rapidly heading in that direction.

I think these progressions of technology are more likely to be like Moore's law: it might be true for a while but eventually it'll peter out. AI itself doesn't understand anything, and there's a limit to human understanding, so technological progression will eventually be self-limiting.

> it's not at all clear that they're really more than a glorified autocorrect.

I use LLMs regularly in my job - many times a day - and I suspect you haven't used them much if you think this.

They're relevant to the singularity discussion though, because they already give a taste of what superhuman intelligence could look like.

ChatGPT, for example, is objectively superhuman in many ways, despite its significant limitations. Once systems like this are more integrated with the outside world and able to learn more directly from feedback, we'll get an even bigger leap forward.

Dismissing this as "glorified autocorrect" is extremely far off base.

> ChatGPT, for example, is objectively superhuman in many ways

But so is a car.

ChatGPT is mostly superhuman in its ability to draw upon enormous numbers of existing sources. It's not superhuman, or even human, in terms of logic, reasoning, or inventing something truly new.