6 ms·

I've replaced 90% of my Google searches with Phind in the last few weeks. My use cases are learning a new API, debugging, generating test cases. It's amazing.

by dalmo3 3y ago

I've replaced 90% of my Google searches with Phind in the last few weeks. My use cases are learning a new API, debugging, generating test cases. It's amazing. Real time saver. Just yesterday it saved me from going down an hour+ rabbit hole due to a cryptic error message. The first solution it gave me didn't work, neither did the second, but I kept pushing and in just a couple of minutes I had it sorted. Having said that, I'm not sure I see the gain with Expert mode yet. After using it for the last couple of days, it's definitely much slower but I couldn't perceive it to be any more accurate. Judging by your example, it looks like the main difference is that the Expert mode search returned a more relevant top result, which then the LLM heavily relied on for its answer. If search results come from bing, can you really credit that answer to Expert mode? PS. You mention launching GPT-4 today, but the Expert Mode toggle has been there for at least a few days, I reckon? Was it not GPT-4 before?

rushingcreek 3y ago
Love to hear it. It's true that for some searches you might not notice a difference, but for complex code examples, reasoning, and debugging Expert mode does seem to be much better. We quietly launched Expert mode a few days ago on our Discord but are now telling the broader HN community about it. We're working on making all of our searches the same quality as Expert mode while being much faster.
- Paul-Craft 3y ago
  I'm definitely giving this a try sometime soon. I had an idea back when it was just GPT-3 out there, to use LLM-generated embeddings as part of a search ranking function. I'm betting that's roughly how Expert mode works, right? Edit: Just had another thought. You could use the output of a normal search algorithm to feed the LLM targeted context, which it could then use to come up with a better answer than it would without the extra background. Yeah, I like that. Although, I will say I asked it about writing a lisp interpreter in Python, because I was just tooling around with such a thing a little while ago for funsies. It essentially pointed me to Peter Norvig's two articles on the subject, which, unfortunately, both feature code that either doesn't run properly or doesn't run right at all. I was disappointed.
  - rushingcreek 3y ago
    We do use the output of a "normal" search algorithm to feed our LLM context :) Did you use Expert mode for your search? Only Expert mode is GPT-4 and its code quality is vastly superior to that of the default mode.
steve-atx-7600 3y ago
google who?
- angelbar 3y ago
  its a verb now, I search at phind.com
Rastonbury 3y ago
I'm a beginner, so I'm unable to tell if it's hallucinating or not. Do you find it hallucinates or is incorrect? I'm wary of noting stuff down and remember wrong things an don't want to drill 2 levels deep for each question
- bentcorner 3y ago
  I just tried it for a problem I solved in Azure Data Explorer and it solved it by making up some APIs that don't exist. It got close to how I solved the problem but cheated even with Expert mode enabled.
  - Rastonbury 3y ago
    Seems like accuracy is the next killer feature for LLM search and teaching, will try again in 6 months
    - LeonenTheDK 3y ago
      What a time to be alive where we likely need wait only a few months for the next big hurdle to be accomplished. Exhilarating and terrifying at the same time.
      - kedean 3y ago
        I dunno about that in this case. The "confidently incorrect" problem seems inherent to the underlying algorithm to me. If it were solved, I suspect that would be a paradigm shift of the sort that happens on the years scale at best.
        mrtranscendence 3y ago
        Yes, the "confidently incorrect" issue will be a tough nut to crack for the current spate of generative text models. LLMs have no ability to analyze a body of text and determine anything about it (e.g. how likely it is to be true); they are clever but at bottom can only extrapolate from patterns found in the training data. If no one has said anything like "X, and I'm 78% certain about it", then it's tough to imagine how an LLM could generate reasonably correct probability estimates.

by dalmo3 3y ago

rushingcreek 3y ago

Love to hear it. It's true that for some searches you might not notice a difference, but for complex code examples, reasoning, and debugging Expert mode does seem to be much better. We quietly launched Expert mode a few days ago on our Discord but are now telling the broader HN community about it. We're working on making all of our searches the same quality as Expert mode while being much faster.

Paul-Craft 3y ago
I'm definitely giving this a try sometime soon. I had an idea back when it was just GPT-3 out there, to use LLM-generated embeddings as part of a search ranking function. I'm betting that's roughly how Expert mode works, right? Edit: Just had another thought. You could use the output of a normal search algorithm to feed the LLM targeted context, which it could then use to come up with a better answer than it would without the extra background. Yeah, I like that. Although, I will say I asked it about writing a lisp interpreter in Python, because I was just tooling around with such a thing a little while ago for funsies. It essentially pointed me to Peter Norvig's two articles on the subject, which, unfortunately, both feature code that either doesn't run properly or doesn't run right at all. I was disappointed.
- rushingcreek 3y ago
  We do use the output of a "normal" search algorithm to feed our LLM context :) Did you use Expert mode for your search? Only Expert mode is GPT-4 and its code quality is vastly superior to that of the default mode.

steve-atx-7600 3y ago

google who?

angelbar 3y ago
its a verb now, I search at phind.com

Rastonbury 3y ago

I'm a beginner, so I'm unable to tell if it's hallucinating or not. Do you find it hallucinates or is incorrect? I'm wary of noting stuff down and remember wrong things an don't want to drill 2 levels deep for each question

bentcorner 3y ago
I just tried it for a problem I solved in Azure Data Explorer and it solved it by making up some APIs that don't exist. It got close to how I solved the problem but cheated even with Expert mode enabled.
- Rastonbury 3y ago
  Seems like accuracy is the next killer feature for LLM search and teaching, will try again in 6 months
  - LeonenTheDK 3y ago
    What a time to be alive where we likely need wait only a few months for the next big hurdle to be accomplished. Exhilarating and terrifying at the same time.
    - kedean 3y ago
      I dunno about that in this case. The "confidently incorrect" problem seems inherent to the underlying algorithm to me. If it were solved, I suspect that would be a paradigm shift of the sort that happens on the years scale at best.
      - mrtranscendence 3y ago
        Yes, the "confidently incorrect" issue will be a tough nut to crack for the current spate of generative text models. LLMs have no ability to analyze a body of text and determine anything about it (e.g. how likely it is to be true); they are clever but at bottom can only extrapolate from patterns found in the training data. If no one has said anything like "X, and I'm 78% certain about it", then it's tough to imagine how an LLM could generate reasonably correct probability estimates.