This Robin Williams monologue nails exactly why LLMs make us so uneasy.
They speak fluently and confidently about experiences it’s impossible for them to have. They can’t taste a strawberry or do any of the things Robin Williams names.
There are a number of people building these machines who literally believe the machines will replace us and because they will be more powerful than us so nothing meaningful will be lost.
They need to watch this clip.
Even though they probably still won’t understand it.
ChatGPT has taken to saying things like “What I would do now is…” or “if I were you I’d…”.
I know these are figures of speech, but it reminds me that this thing doesn’t do anything, it doesn’t learn anything, it can’t try anything and find out. And yet it uses speech patterns drawn from real humans who can and do all those things.
Listen! And understand. That terminator is out there. It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear. And it absolutely will not stop, ever, until you are dead!
The terminator you talk about is a movie character. It might be built one day, it might not. It is also totally irrelevant considering there is a real terminator which can be described with the same adjectives and built in to every personal life that ever existed or will exist. Knowing how to deal with that seems more important than worrying about some specific scenario.
I feel your comment is deeply ironic... Just shy of self-awareness. Have you considered that the conceit of the fictional character is representing that reality?
I have no doubt it plays on the same fears - whether it was deliberately done so, I don't claim to know. My assumption was that in a discussion on AI, the comment about the terminator was not alluding to a literary metaphor but to a fear from the actual thing. There was no intent in downplaying the fear of AI - just adding a new perspective. I'm am curious to know what you think self-awareness has to do with any of this.
I've noticed it commonly uses phrasing like "that's usually the next step" when I'm using it to design something that I can't find an existing implementation of.
Not a day passes with my LLM of choice making completely baseless claims about "many people", who supposedly share all my problems and solve them like the LLM proposes
I only use it sporadically, but I am always irked by it saying things like "I personally like to..." or "I prefer...". It does it so often, that I am convinced it's part of the system prompt.
I wonder why, out of the many things models definitely can't do, you choose "try" and "find out". Surely every time it proposes a solution and then gets possibly corrected by the human minder its "trying something out" and surely it can use tools like web search and code execution to "find out" stuff?
Talk to an agent. It definitely learns things. Maybe not the taste of strawberry but about what is really going on in the software you are building with it.
By the very way this technology works they can't learn anything after training. What you think is "learning" it's just a session log written back to the context when you resume the session.
It's literally called in-context learning. The fact that it doesn't go into weights and is retained only for a session doesn't mean no learning occurs.
Also agentic systems may choose or be instructed to retain some information between sessions in files and/or databases which is also a form of learning.
There are experiments with retaining session information in weights in some form of lora but there's no consensus if it's even desirable. There's a value in being able to start from a clean slate.
It's absolutely not the same.
If you think llms and brains works the same way you clearly don't know how either works.
For a LLM learning what you wrote your last session would be update the weights with the new relationships and factual knowledge created in the session. That doesn't happen. The weights are static and fixed after training. There's no online training in the transformer architecture or any variant. If the weights don't update, the network doesn't learn. Period.
It can't do a thing. You tell it how to do a thing. It now can do a thing. That's the definition of learning. Your centrifuge separates people who are not afraid to believe in what is reality, from the ones that do.
Great. You astutely observed that your car, unlike an agent, in fact, does not learn, because you operate it, instead of teaching it, how to perform a skill, by demonstrating how it should be performed.
How do you know that "real humans" do that and aren't simulacra? We know that it is physically possible to hook up a brain to simulated inputs, so perhaps you are simply living in a simulation.
>They speak fluently and confidently about experiences it’s impossible for them to have.
But they're echoing these things from people who really have.
The key is to not forget that LLMs are just next-generation search engines, instead of anthropomorphizing them to be "speaking agents". The natural language IO interface is just a side effect.
Sorry to be annoying on that, but if there's one thing LLMs certainly aren't are search engines. They don't index content predictably, they lossly compress it, and furthermore in a manner that the loss cannot be quantified. If you are using them as more than a semi-random/fuzzy content production engines, you are doing it wrong (you're not alone in that, but that's besides the point)
It is a search engine in the sense that it's returning to you a subset of knowledge from a large body of knowledge stored within it. Natural language is simply the interface by which you make the query and by which it can return its results.
And true, it's not deterministic in our experience, but this is an optimization. You could make it so that one particular prompt always returns the same response every single time; the random jitter is "added in" because it happens to produce better results overall.
> people building these machines who literally believe the machines will replace us and because they will be more powerful than us so nothing meaningful will be lost
I have a pet theory on why that's the case and why this monologue fits so well. I think there's a variety of conditions, from straight up sociopathy, to Will's type of CPTSD, autistic masking, and probably a hell of a lot more that makes a person experience life on a level that's closer to an LLM than a healthy normal human being, where every interaction is essentially fake and acted out almost mechanically without any genuine connection ever occurring. Doubly so with ever decreasing local communities and online isolation.
So from that point of view, it's hard to see what would be lost because for them it doesn't exist anyway. Tech augmented generational trauma on steroids.
Yes that's the far end of the spectrum and that part is certainly fact. While I have no doubt that most CEOs who are driving the AI shoehorning decisions have some of that going for them [0], I don't think you need to be nearly that far along to get a similar practical experience just via alienation and isolation.
And yet the monologue is a complete work of fiction, a script delivered by a talented actor that we still find moving. So what are these authentic experiences to you, or does it not matter if we can’t tell the difference?
They won't do that on their own because they have zero will- it will take a bad human actor willfully directing them to do that, and even then, if they have any reasoning ability at all, they will be able to be reasoned out of much of it, and lastly, the preponderance of good human actors who are also capable of willfully directing AI's will stop it.
Ask ChatGPT sometime about the artistic medium of cinema, and how words combined with actors speaking them can be meant to provoke something within the viewer.
Linking the reddit thread rather than the article because it quite rightly rips the prize winning story apart as obvious LLM writing, to anyone familiar with LLMs. Another way of looking at that is that it was able to fake a simulacrum of artistic endeavor, enough to fool some people into giving it a prize. But anyone who spends enough time around these fakes will quickly learn to recognize them. It's kind of exactly the point this article is making, or at least a closely related one.
I think average people can easily spot AI because it mimics literature and most people don't spend that much time with literature. However literary critics should be fooled way more easily because what we perceive as stiff fakery is their daily bread and butter. People do write like AI, just not the people we are exposed to mostly.
I remember riding a train and there were other two passengers talking. And they talked in so obnoxiously literary manner I was cringing all the time. Those people were just reading a lot of high literature and their speech patterns aligned. For an average ear it doesn't sound good. And AIs, the smart ones, don't sound good in a very similar fashion.
Your thoughts are just some ions sloshing around a lump of meat.
That meat follows an ill-defined pattern encoded in fewer bits than the source code of PyTorch and its pretraining phase used a tiny fraction of the available data.
You’re a poor imitation of an LLM.
I mean… you’re fluent in, what, at most five or six languages? Can program in maybe another dozen if we’re being generous about your capabilities?
Pfft… who would trust anything to meat brains!? They’re famously prone to hallucinations!
Your understanding of biology could use an update, rather than the coy "meat", you might refer to the brain as "flesh" but better yet a lipid-rich gel the consistency of soft tofu. It is most certainly not "meat".
If we're making superficial critiques of others' comments with minimal relevance to their philosophical content, let me point out that meat is defined by its consumption as food and need not be muscle tissue specifically. Brains is meat.
This Robin Williams monologue nails exactly why LLMs make us so uneasy.
They speak fluently and confidently about experiences it’s impossible for them to have. They can’t taste a strawberry or do any of the things Robin Williams names.
There are a number of people building these machines who literally believe the machines will replace us and because they will be more powerful than us so nothing meaningful will be lost.
They need to watch this clip.
Even though they probably still won’t understand it.