Maybe so, but we already have an example of a generally intelligent system that outperforms our current AI models in its cognitive capabilities while using orders of magnitude less power and memory: the human brain. That alone suggests our current brute‑force approach probably won’t be the path a true AGI takes. It’s entirely conceivable that such a system improves through optimization - getting better while using less power, at least in the beginning.

I personally think the whole concept of AGI is a mirage. In reality, a truly generally intelligent system would almost immediately be superhuman in its capabilities. Even if it were no “smarter” than a human, it could still process information at a vastly higher speed and solve in minutes what would take a team of scientists years or even decades.

And the moment it hits “human level” in coding ability, it starts improving itself - building a slightly better version, which builds an even better version, and so on. I just don’t see any plausible scenario where we create an AI that stays at human-level intelligence. It either stalls far short of that, or it blows right past it.

If AI ends up destroying us, I’d say it’s unlikely to be because it hates us or wants to destroy us per se - more likely it just treats us the way we treat ants. We don’t usually go out of our way to wipe out ant colonies, but if there’s an anthill where we’re putting up a house, we don’t think twice about bulldozing it. Even in the cartoonish “paperclip maximizer” thought experiment, the end of humanity isn’t caused by a malicious AI - it’s caused by a misaligned one.

That would by definition mean it’s not superintelligent.

Superintelligence doesn’t imply ethics. It could just as easily be a completely unconscious system that’s simply very, very good at crunching data.

If you’re genuinely interested in what “artificial superintelligence” (ASI) means, you can just look it up. Zuckerberg didn’t invent the term - it’s been around for decades, popularized lately by Nick Bostrom’s book Superintelligence.

The usual framing goes like this: Artificial General Intelligence (AGI) is an AI system with human-level intelligence. Push it beyond human level and you’re talking about Artificial Superintelligence - an AI with cognitive abilities that surpass our own. Nothing mysterious about it.

Judging by the comments here I’m getting the impression that people would like to rather provide a selfie or ID.

No other than it’s geographically closer to my actual location so I thought the speed would be faster.

EU is about to do the exact same thing. Norway is the place to be. That’s where I went - at least according to my ip address.

FUD has nothing to do with what this is about.

And nothing of value was lost.

Sure, if privacy is worth nothing to you but I wouldn’t speak for the rest of the UK and EU.

My feed right now.

It’s actually the opposite of a very specific definition - it’s an extremely broad one. “AI” is the parent category that contains all the different subcategories, from the chess opponent on an old Atari console all the way up to a hypothetical Artificial Superintelligence, even though those systems couldn’t be more different from one another.

It’s a system designed to generate natural-sounding language, not to provide factual information. Complaining that it sometimes gets facts wrong is like saying a calculator is “stupid” because it can’t write text. How could it? That was never what it was built for. You’re expecting general intelligence from a narrowly intelligent system. That’s not a failure on the LLM’s part - it’s a failure of your expectations.

I don’t think you even know what you’re talking about.

You can define intelligence however you like, but if you come into a discussion using your own private definitions, all you get is people talking past each other and thinking they’re disagreeing when they’re not. Terms like this have a technical meaning for a reason. Sure, you can simplify things in a one-on-one conversation with someone who doesn’t know the jargon - but dragging those made-up definitions into an online discussion just muddies the water.

The correct term here is “AI,” and it doesn’t somehow skip over the word “artificial.” What exactly do you think AI stands for? The fact that normies don’t understand what AI actually means and assume it implies general intelligence doesn’t suddenly make LLMs “not AI” - it just means normies don’t know what they’re talking about either.

And for the record, the term is Artificial General Intelligence (AGI), not GAI.

Claims like this just create more confusion and lead to people saying things like “LLMs aren’t AI.”

LLMs are intelligent - just not in the way people think.

Their intelligence lies in their ability to generate natural-sounding language, and at that they’re extremely good. Expecting them to consistently output factual information isn’t a failure of the LLM - it’s a failure of the user’s expectations. LLMs are so good at generating text, and so often happen to be correct, that people start expecting general intelligence from them. But that’s never what they were designed to do.

There are plenty of similarities in the output of both the human brain and LLMs, but overall they’re very different. Unlike LLMs, the human brain is generally intelligent - it can adapt to a huge variety of cognitive tasks. LLMs, on the other hand, can only do one thing: generate language. It’s tempting to anthropomorphize systems like ChatGPT because of how competent they seem, but there’s no actual thinking going on. It’s just generating language based on patterns and probabilities.

Large language models aren’t designed to be knowledge machines - they’re designed to generate natural-sounding language, nothing more. The fact that they ever get things right is just a byproduct of their training data containing a lot of correct information. These systems aren’t generally intelligent, and people need to stop treating them as if they are. Complaining that an LLM gives out wrong information isn’t a failure of the model itself - it’s a mismatch of expectations.

Way to move the goalposts.

If you take that question seriously for a second - AlphaFold doesn’t spew chemicals or drain lakes. It’s a piece of software that runs on GPUs in a data center. The environmental cost is just the electricity it uses during training and prediction.

Now compare that to the way protein structures were solved before: years of wet lab work with X‑ray crystallography or cryo‑EM, running giant instruments, burning through reagents, and literally consuming tons of chemicals and water in the process. AlphaFold collapses that into a few megawatt‑hours of compute and spits out a 3D structure in hours instead of years.

So if the concern is environmental footprint, the AI way is dramatically cleaner than the old human‑only way.

Artificial intelligence isn’t designed to maximize human fulfillment. It’s built to minimize human suffering.

What it cannot do is answer the fundamental questions that have always defined human existence: Who am I? Why am I here? What should I do with my finite time on Earth?

Expecting machines to resolve existential questions is like expecting a calculator to write poetry. We’re demanding the wrong function from the right tool.

Pretty weird statements. There’s no such thing as just “AI” - they should be more specific. LLMs aren’t designed to maximize human fulfillment or minimize suffering. They’re designed to generate natural-sounding language. If they’re talking about AGI, then that’s not designed for any one thing - it’s designed for everything.

Comparing AGI to a calculator makes no sense. A calculator is built for a single, narrow task. AGI, by definition, can adapt to any task. If a question has an answer, an AGI has a far better chance of figuring it out than a human - and I’d argue that’s true even if the AGI itself isn’t conscious.