Post

The AI bros (who clearly don't understand AI) in the comments are hilarious. "ask stupid questions, get stupid answers". LLMs have no concept of how intelligent a question is, and they certainly were not designed to deliberately give stupid responses if they deem your question too stupid. 1/?

What's most likely happening is easier to understand if you mentally model LLMs as glorified search engines. They are matching your search query (prompt) with some text they found on the internet. Internally it's a bit more complex than that, but much closer to Google than to a thinking machine. 2/?

Essentially, it doesn't have an answer for whether you should walk or drive to a car wash, as it's not a commonly asked question (because it's stupid). Just like if you Google searched the question, you wouldn't find a good result because it's not a common question. 3/?

But the question of whether you should walk or drive a given distance is common, so it basically just ignores the car wash part and answers the question of walking or driving in a more generic sense. The problem isn't that someone is going to ask the LLM this exact question, it's the handling. 4/?

It clearly can't infer from a simple one sentence question that I might need my car to wash my car, because it isn't intelligent, doesn't possess logic, and can't reason. When you ask it anything complex, it's coming across 10s, 100s, or 1000s of points where it has to infer context. 5/?

The reason it appears to do well at much more complex problems is because there are "search results" for that "query". It can explain general relativity, because it has data from millions of people explaining general relativity, not because it understands any of the underlying concepts. 6/?

So, when you use LLMs as thinking machines, you will run into variants of a core problem: It isn't reasoning. It's just a fancy bot that copy & pastes then rewords data written by humans. Which means: 1. It's implicitly capped by its human produced-training data, which is largely garbage. 7/?

2. if you try to use it for a task that diverges too far from something it has in its training data, it will shit the bed (as is evident by the car wash question, which any intelligent being could easily answer). 8/?

3. Even when working on tasks for which the LLM has comprehensive data on, it's having to infer context constantly, and every time it mischaracterizes or drops some context, it goes further off the rails. 9/?

Just as I would not think to explain to someone that they need to go to the car wash WITH THEIR CAR in order to wash it (because most people aren't that stupid). Most human produced text forgoes explaining a lot of obvious context, as it's not needed. Which is a problem if LLMs can't infer it. 10/?

So the OP is just a really obvious example of how LLMs fundamentally do not understand things, thus miss even the most obvious context. Yet people are happy to pretend that when doing complex tasks, they just magically perfectly understand every bit of implied context encountered along the way. 11/?

In many instances it's not about "them" being able to understand, it's about us being able to understand them. Ignore that at our own risk. youtu.be/JKk77rzOL34?...

If you think it's a waste of time to listen to it just jump to the 13:00 minute mark. That'll probably get your attention!

Isn't the onus on the person making this argument to finish the sentence with "as opposed to humans which are something else" and explain how we are different? Or what constitutes human understanding that is somehow different to a glorified predictive search engine?

humans demonstrate teleological thought. a human can identify the necessity of bringing your car at a car wash because it understands what a car is and what a car wash is

What people are doing is trying to confirm or falsify what you're saying and being unable to. There isn't really a consensus definition of "think," "know," "understand," or "reason," so it's not really meaningful to make a statement like this.

It seems to me to be bait more than anything else.

We know a rock doesn't think and we know humans do. We're not at the point where there is any doubt if this thinks or not.

Whilst I'm not arguing about AI here, agree with you, it does not think, we don't actually have any scientific definitions for consciousness, intelligence or actually even 'alive'.

Great don’t use them.

Saved

Computers do exactly what they're told to, nothing more, nothing less. As a 90s compsci grad I feel like I'm yelling into the void half the time.

When I was in my first computer class in 1979, my instructor told us that a computer was the dumbest thing on earth. It does exactly what you tell it to do, even if your instructions in retrospect were bad.

LLM AI is not even a search engine. Search engines often point you in the right direction. LLM AI is just a parrot - a parrot with a memory the size of a planet and less brain power than a slug. LLM AI has no common sense. The closest LLM answer to the correct answer is often the exact opposite.

Let us assume that an LLM has been trained on the entire corpus of human knowledge since the dawn of time. Let us also generously assume that every sentence it is given to absorb is 100% factually accurate. Let us then allow the LLM to assemble new sentences using weighted random heuristics out of

the material it has been given. What is the probability that the sentences the LLM emits will also be 100% factually accurate? Oh, it's zero, isn't it?

If true, not bad for a search engine. blog.google/innovation-a... Also, you might want to check out on what François Chollet is working on. ndea.com These concepts go far beyond simple probabilistic or semantic search.

Chollet has said for a while that the limitations of LLMs will not lead to AGI, so I don't think his view is in opposition to the OP. forum.effectivealtruism.org/posts/MGpJpN...

OP is not stating AGI is not there yet. He says it's all useless slop machines. Mr. Chollet might oppose this while knowing where current limitations are. The topic is more complex than total slop or AGI.

If I click on this and it turns out to be bullshit, what do I get in return for my time? Because my decades of experience on the web tells me people will link to bullshit and expect me to read it.