I’ll explain for you, because there’s a lot of misinformation around.
What is being called AI these days is various companies’ version of what’s called an LLM – Large Language Model. Put simply, an LLM is a very sophisticated piece of software that takes what is asked of it to determine what is statistically the most likely sequence of words to follow as an answer.
This means you can ask a question the way you’d ask a human, and the way it answers will closely mirror how a person would answer (as opposed to stuff like Google Assistant or Siri, where you need to ask a question a specific way to get a decent answer).
Note, however, that at no point did I say that an LLM is accurate. This is the fatal issue that is never included by proponents of this kind of AI. They don’t have any mechanism to retrieve information, or verify the truthfulness of the answers given. You wind up seeing a lot of answers from this kind of AI that is either partially or completely wrong.
My favorite example is the result you get when googling “african countries that start with the letter K”. Someone posted the answer they got from an LLM to a forum online, which said that there is no country, and that became the top google result…despite the fact that Kenya obviously exists and starts with the letter K.
Essentially, LLMs are really fascinating in how well they approximate human speech – but they have absolutely no intelligence behind them. Proponents of this tech as AI either ignore this, or outright lie about it. As a result, a lot of companies have started using this tech to replace their support teams and/or the search functionality of their websites. I’m sure you can imagine the negative effects this has caused.
If you want to know the state of “AI” right now, just try calling customer service or talking to a ChatBot for any company. It’s incredibly sh*tty.
But for real, if you want to know the state of AI, go to Hugging Face.
Mind explaining to a tech layperson why they’re bad?
I’ll explain for you, because there’s a lot of misinformation around.
What is being called AI these days is various companies’ version of what’s called an LLM – Large Language Model. Put simply, an LLM is a very sophisticated piece of software that takes what is asked of it to determine what is statistically the most likely sequence of words to follow as an answer.
This means you can ask a question the way you’d ask a human, and the way it answers will closely mirror how a person would answer (as opposed to stuff like Google Assistant or Siri, where you need to ask a question a specific way to get a decent answer).
Note, however, that at no point did I say that an LLM is accurate. This is the fatal issue that is never included by proponents of this kind of AI. They don’t have any mechanism to retrieve information, or verify the truthfulness of the answers given. You wind up seeing a lot of answers from this kind of AI that is either partially or completely wrong.
My favorite example is the result you get when googling “african countries that start with the letter K”. Someone posted the answer they got from an LLM to a forum online, which said that there is no country, and that became the top google result…despite the fact that Kenya obviously exists and starts with the letter K.
Essentially, LLMs are really fascinating in how well they approximate human speech – but they have absolutely no intelligence behind them. Proponents of this tech as AI either ignore this, or outright lie about it. As a result, a lot of companies have started using this tech to replace their support teams and/or the search functionality of their websites. I’m sure you can imagine the negative effects this has caused.
Oh I meant Hugging Face in particular.
Still, I appreciate you taking the time to lay out that explanation!
Oh whoops! 😬 I hope you get the answer you needed!
I think they’re being more literal. All the latest open source AI models get posted on Hugging Face.
deleted by creator