zztong
Hunter
I just want to note I don't begrudge anyone for using an AI. I use several of them daily as in my profession (cybersecurity professor) its usage is surging and it's usage by my students is surging. I'm sure Perplexity is a great tool. I certainly intend to try it out further.
From my own experiences: AI can be pretty handy in writing code; but still makes mistakes. AI can be pretty good as a Wikipedia-like source of information. When it struggles with sourcing information it hallucinates. When it leans on Reddit, it can't tell good information from bad. (Come to think of it, that's a lovely summary of Reddit.) AI's can become a kind of echo-chamber if you engage in long conversations with it, telling you what you want to hear.
I have colleagues attempting to make more reliable chatbots. They spend a lot of time giving it reliable sources of information and then even more time trying to get it to say "I don't know" if it can't come up with an answer from the trusted sources.
Among my students, many get trapped when they ask AI for an answer and AI pulls information from an outdated standard. I have a folder full of hallucinations from my own usage. One AI gave me a summary of the "Walter Payton Cybersecurity Act" which doesn't exist.
I look forward to using it. Claude is often touted as the best at writing code. As one guy in a seminar put it, ChatGPT and CoPilot are drunk junior programmers, while Claud is a weak senior programmer. I like ChatGPT's threads and it looks like Perplexity has the same feature.
From my own experiences: AI can be pretty handy in writing code; but still makes mistakes. AI can be pretty good as a Wikipedia-like source of information. When it struggles with sourcing information it hallucinates. When it leans on Reddit, it can't tell good information from bad. (Come to think of it, that's a lovely summary of Reddit.) AI's can become a kind of echo-chamber if you engage in long conversations with it, telling you what you want to hear.
I have colleagues attempting to make more reliable chatbots. They spend a lot of time giving it reliable sources of information and then even more time trying to get it to say "I don't know" if it can't come up with an answer from the trusted sources.
Among my students, many get trapped when they ask AI for an answer and AI pulls information from an outdated standard. I have a folder full of hallucinations from my own usage. One AI gave me a summary of the "Walter Payton Cybersecurity Act" which doesn't exist.
Also IMHO Perplexity has been the most useful, well worth looking into.
I look forward to using it. Claude is often touted as the best at writing code. As one guy in a seminar put it, ChatGPT and CoPilot are drunk junior programmers, while Claud is a weak senior programmer. I like ChatGPT's threads and it looks like Perplexity has the same feature.