5 ms·

LLMs sound so confident and expert-like that people going to follow their advice, no matter how much you warn them. I tried out https://www.perplexity.ai/ http

by nabla9 2y ago

LLMs sound so confident and expert-like that people going to follow their advice, no matter how much you warn them. I tried out https://www.perplexity.ai/ https://www.perplexity.ai/ because their model give plenty of references for the claims it makes. But when you check the references, they often don't say what the model claims. But no matter, it sounds even more confident because it provides sources. LLM's are near perfect bullshitters using Harry Frankfurt's definition: "When an honest man speaks, he says only what he believes to be true; and for the liar, it is correspondingly indispensable that he considers his statements to be false. For the bullshitter, however, all these bets are off: he is neither on the side of the true nor on the side of the false. His eye is not on the facts at all, as the eyes of the honest man and of the liar are, except insofar as they may be pertinent to his interest in getting away with what he says. He does not care whether the things he says describe reality correctly. He just picks them out, or makes them up, to suit his purpose."

barfbagginus 2y ago
You're using low grade AI, and as a result you might have an exaggerated view of how much AI bullshits. Perplexity, especially, is nearly unusable. Even though it has references, it's still a GPT3.5 equivalent model, so it will hallucinate about the references even if they're right in front of it. When you correct one hallucination, it'll thank you, then immediately make another one. Every conversation is a tragic comedy of errors. It's frustrating and upsetting for any real work. But higher grade AI can be a useful coworker. You will have to pay for pay for Gemini 1.5, GPT4, or Claude 3 Opus, and go through the learnprompting.org lessons, while reading the papers referenced in the lessons. These AIs can usefully critique and interact with my work and references. They can bullshit, but rarely do so, given good references. When they do BS, I can politely call them on it and they fix it and go back to solid reasoning. And even better, they can often critique and call out MY bullshit. I have long conversations with the AI. They start with a big problem that is stressing me. And they usually end with an amazing solution that is easy for me to execute. Usually I could not have easily formulated on my own, and might even have gotten stuck or distracted on the task. It's good stuff - don't footgun yourself by cheaping out on good AI, or not reading the promoting papers. All three of those AIs cost about 20 a month. But if you want AI that actually works with you, it's more than worth it.

by nabla9 2y ago

barfbagginus 2y ago

You're using low grade AI, and as a result you might have an exaggerated view of how much AI bullshits. Perplexity, especially, is nearly unusable. Even though it has references, it's still a GPT3.5 equivalent model, so it will hallucinate about the references even if they're right in front of it. When you correct one hallucination, it'll thank you, then immediately make another one. Every conversation is a tragic comedy of errors. It's frustrating and upsetting for any real work. But higher grade AI can be a useful coworker. You will have to pay for pay for Gemini 1.5, GPT4, or Claude 3 Opus, and go through the learnprompting.org lessons, while reading the papers referenced in the lessons. These AIs can usefully critique and interact with my work and references. They can bullshit, but rarely do so, given good references. When they do BS, I can politely call them on it and they fix it and go back to solid reasoning. And even better, they can often critique and call out MY bullshit. I have long conversations with the AI. They start with a big problem that is stressing me. And they usually end with an amazing solution that is easy for me to execute. Usually I could not have easily formulated on my own, and might even have gotten stuck or distracted on the task. It's good stuff - don't footgun yourself by cheaping out on good AI, or not reading the promoting papers. All three of those AIs cost about 20 a month. But if you want AI that actually works with you, it's more than worth it.