7 ms·

I've been using the claude model over the past months and I have to say I usually prefer it over ChatGPT. Some things that impressed me particular are: * It c

• lucgagan 3y ago
  As a counter point, I've been actively assessing feasibility of migrating my experiments to Claude with not much success. Trying to migrate prompts from GPT 4 to Claude often yields half-baked results. Claude seems to require a lot more handholding to get the desired result.
  • visarga 3y ago
    In my information extraction prompts Claude is about as good as GPT-3.5 but slower. Its only advantage over OpenAI is context length. GPT-4 is better, 3.5 is faster. But I am using it often to chat with PDFs and it works great for that. I would not want to give this ability up.
    • baq 3y ago
      > But I am using it often to chat with PDFs and it works great for that. I would not want to give this ability up. Side note: this would get you a crackpot moniker a year ago. Chat with pdfs? Dude what are you smoking I need some too. Perfectly normal today, in fact if you aren’t doing that you’re falling behind.
• amanzi 3y ago
  You didn't say which version of GPT you prefer it over. My experience of Claude hasn't been good at all - I get obvious hallucinations all the time. Whereas I get pretty consistent accurate results with GPT-4. I just tried again now and asked it about HTMX and Django since I had just been reading an article about it. Claude invented a package called `htmx.django` and a decorator that could be imported called `@hx_request`. This is typical of my experience with Claude.
  • jeffreygoesto 3y ago
    Why didn't you let it implement it and start a new project?
  • edgyquant 3y ago
    I’ve been using Claude for months now, as has my dev team and a number of other people I know, and we don’t have this problem at all. In my experience is hallucinates less. This is compared to ChatGPT with GPT4. Anecdotal evidence is worthless it seem.
    • jazzyjackson 3y ago
      hallucination is stochastic. Out of 1000 people flipping coins over and over again, on average people get a mix of hits and misses - there will always be someone who just get 10 heads in a row and someone else with 10 tails in a row, this corresponds to some users being wowed by the superintelligence nailing the answer every time and other users getting made up citation after made up citation.
  • shmatt 3y ago
    Claude is bad at code and great at everything else, I thought that was pretty common knowledge by now. I second OP where I prefer using it over OpenAI models these days
  • inciampati 3y ago
    It won't hallucinate almost anything if you give it the entire documentation for the libraries you want to use. It can ingest many hundreds of pages of documentation without problem. Used like this it is far more powerful than GPT-4 or any existing model.
    • SubiculumCode 3y ago
      i know some people say its extended context is more hackery than not, but I've found better success with Claude in roughing out some paragraphs before editing when i provide it a lot of context
• airstrike 3y ago
  ChatGPT 3.5 or 4? The difference between them is much more massive than "0.5" appears to suggest
• m3kw9 3y ago
  When using stuff like this you let the cream of top surface then use it. It will surface when enough people/publication give praise, otherwise you waste time evaluating too many models, it’s not a fun task as they can give you illusions of being great at some but end up with overall worse. Right now gpt4 is king and haven’t been any consistent chatter saying otherwise