I have many conversations with people about Large Language Models like ChatGPT and Copilot. The idea that “it makes convincing sentences, but it doesn’t know what it’s talking about” is a difficult concept to convey or wrap your head around. Because the sentences are so convincing.
Any good examples on how to explain this in simple terms?
Edit:some good answers already! I find especially that the emotional barrier is difficult to break. If an AI says something malicious, our brain immediatly jumps to “it has intent”. How can we explain this away?
It’s a really well-trained parrot. It responds to what you say, and then it responds to what it hears itself say.
But despite knowing which sounds go together based on which sounds it heard, it doesn’t actually speak English.
idea that “it makes convincing sentences, but it doesn’t know what it’s talking about”
Like a teenager who has come into a new group and is now trying so hard to fit in :-)
I commented something similair on another post, but this is exactly why I find this phenomenon so hard to describe.
A teenager in a new group still has some understanding and has a mind. It knows many of the meaning of the words that are said. Sure, some catchphrases might be new, but general topics shouldn’t be too hard to follow.
This is nothing like genAI. GenAI doesn’t know anything at all. It has (simplified) a list of words that somehow are connected to eachother. But AI has no meaning of a wheel, what round is, what rolling is, what rubber is, what an axle is. NO understanding. Just words that happened to describe all of it. For us humans it is so difficult to understand that something uses language without knowing ANY of the meaning.
How can we describe this so our brains make sense that you can have language without understanding? The Chinese Room experiment comes close, but is quite complicated to explain as well I think.
It’s just fancy predictive text like while texting on your phone. It guesses what the next word should be for a lot more complex topics.
I am an LLM researcher at MIT, and hopefully this will help.
As others have answered, LLMs have only learned the ability to autocomplete given some input, known as the prompt. Functionally, the model is strictly predicting the probability of the next word+, called tokens, with some randomness injected so the output isn’t exactly the same for any given prompt.
The probability of the next word comes from what was in the model’s training data, in combination with a very complex mathematical method to compute the impact of all previous words with every other previous word and with the new predicted word, called self-attention, but you can think of this like a computed relatedness factor.
This relatedness factor is very computationally expensive and grows exponentially, so models are limited by how many previous words can be used to compute relatedness. This limitation is called the Context Window. The recent breakthroughs in LLMs come from the use of very large context windows to learn the relationships of as many words as possible.
This process of predicting the next word is repeated iteratively until a special stop token is generated, which tells the model go stop generating more words. So literally, the models builds entire responses one word at a time from left to right.
Because all future words are predicated on the previously stated words in either the prompt or subsequent generated words, it becomes impossible to apply even the most basic logical concepts, unless all the components required are present in the prompt or have somehow serendipitously been stated by the model in its generated response.
This is also why LLMs tend to work better when you ask them to work out all the steps of a problem instead of jumping to a conclusion, and why the best models tend to rely on extremely verbose answers to give you the simple piece of information you were looking for.
From this fundamental understanding, hopefully you can now reason the LLM limitations in factual understanding as well. For instance, if a given fact was never mentioned in the training data, or an answer simply doesn’t exist, the model will make it up, inferring the next most likely word to create a plausible sounding statement. Essentially, the model has been faking language understanding so much, that even when the model has no factual basis for an answer, it can easily trick a unwitting human into believing the answer to be correct.
—-
+more specifically these words are tokens which usually contain some smaller part of a word. For instance, understand and able would be represented as two tokens that when put together would become the word understandable.
Imagine that you have a random group of people waiting in line at your desk. You have each one read the prompt, and the response so far, and then add a word themself. Then they leave and the next person in line comes and does it.
This is why “why did you say ?” questions are nonsensical to AI. The code answering it is not the code that wrote it and there is no communication coordination or anything between the different word answerers.
Ok, I like this description a lot actually, it’s a very quick and effective way to explain the effects of no backtracking. A lot of the answers here are either too reductive or too technical to actually make this behavior understandable to a layman. “It just predicts the next word” is easy to forget when the thing makes it so easy to be anthropomorphized subconsciously.
