I gave LLMs a bit of a go but a couple of occasions wasted a bunch of time by giving incorrect answers, so my trust levels are pretty low.
Ask HN: Do you still use search engines?
311–320 of 609 posts
Re: Ask HN: Do you still use search engines?
#312Earlier quoted context omitted.
This is the biggest argument in favor of subscription-based funding models for search I've heard. (Kagi et al.) Ad-sponsored models are going to be dead as soon as people realize they can't trust output. And because the entire benefit to LLM search is the convenience of removing a human-in-the-loop step (scanning the search results), there won't be a clear insertion/distinction point for ads without poisoning the ent…
There is no exclusive-or between business models, there is an inclusive-or. Over time subscription models will converge to subscription with advertisements. Like newspapers did.
And streaming services. People really thought streaming would be the end of TV commercials. Ha!
Re: Ask HN: Do you still use search engines?
#313Kagi is a very good alternative to google. When you're actually doing some research and have an exhaustive look at search results, kagi provides much more detailed results than google.
I'd rank kagi > chatgpt > google any day.
Re: Ask HN: Do you still use search engines?
#314Earlier quoted context omitted.
IMO an example of a good use case for an LLM, which would be otherwise very hard to search for, is clarifying vague technical concepts. Just yesterday I was trying to remember the name of a vague concept I’d forgotten, with my overall question being: “Is there a technical term in biology for the equilibrium that occurs between plant species producing defensive toxins, and toxin resistance in the insect species that f…
that sounds like a really great example of..... searching through vector embedding. I don't think the LLM part was technically necessary.
And the LLM won't be necessary either for the common search case of "dc plane crash" or whatever, but increasingly that will be asked of general assistance agent, and it will dispatch an internet search of some kind, and return the result. Even if it provides 0 additional benefit over google, switching the default search location might still occur, especially if it does provide some benefits (such as summarization)
Re: Ask HN: Do you still use search engines?
#315Earlier quoted context omitted.
Unfortunately DDG still enshittifies their results with AI garbage as much as they can. Both with the intrusive AI blob at the top and replacing the page summaries with dreadful AI goobleygook.
You can disable the AI summary thing: https://duckduckgo.com/duckduckgo-help-pages/duckai/how-to-h...
Other way around. That disables the AI chatbot monstrosity at the top of the page. But only if you're logged in and/or don't clear your cookies.
The AI word salad summaries for each individual page have no toggle (unless you count !g).
Re: Ask HN: Do you still use search engines?
#316Daily. On top of general search engines, I have a bunch of custom ones added to Firefox so I can skip the Google/Bing/Yandex step.
Re: Ask HN: Do you still use search engines?
#317AFAICT ChatGPT is mostly useless and can't be trusted to answer questions accurately. So no, mostly all search engines. To be honesty I'm surprised anybody uses it for anything other than trivial uses.
Are you using a paid version? Do you use web-search? And have you tried alternatives like Claude?
I've mostly switched to using Claude these days, with MCPs for websearch and fetching specific remote or local files. It answers questions generally very accurately (from the source documents it identifies) and includes citations.
I've found that people that haven't really tried the latest models, and just rely on whatever knowledge is in the model training are really missing out on the potential power. GPT4o+ and equivalent models really changed the game. And using tools to do a search, or pull in your code, or run a db query or whatever enables them to either synthesize information or generate context relevant material. Not perfect for everything, but much better than a year ago, or what people are doing with the free systems.
Re: Ask HN: Do you still use search engines?
#318Search is primarily a portal - you know a particular resource exists, you just don't know its exact URL. You hear about this new programming language called "Frob", and you assume it must have a website. So you google "Frob language". You hear that there was a plane crash in DC, and assume (CNN/AP/your_favorite_news_site) has almost certainly written an article about it. You google "DC plane crash." LLMs aren't ever…
IMO an example of a good use case for an LLM, which would be otherwise very hard to search for, is clarifying vague technical concepts. Just yesterday I was trying to remember the name of a vague concept I’d forgotten, with my overall question being: “Is there a technical term in biology for the equilibrium that occurs between plant species producing defensive toxins, and toxin resistance in the insect species that f…
Same. This is one of the few uses for LLMs that I actually find useful and that I trust.
They’re very helpful for helping me ask more refined questions by getting the terminology correct.
Re: Ask HN: Do you still use search engines?
#319Earlier quoted context omitted.
that sounds like a really great example of..... searching through vector embedding. I don't think the LLM part was technically necessary.
Okay, so how does the average person search through vector embedding? I would like to try this out.
I don't think this is _literally_ a search through vector embeddings.
LLMs store embeddings of individual tokens (usually parts of words), so a result of an actual search will be top-k embeddings and the corresponding tokens, similar to the output of a Google search. You could extract the initial matrix of embeddings from some open-weights model and find tokens closest to your query. However, it's not clear why do this. OP got coherent text, so that's not search.
It's _similar_, though, because attention in LLMs basically looks for most similar tokens. So to answer the question about the term, the LLM had to create a stream of tokens that's semantically closest to the given description. Well, this is somewhat like a search, but it's not exactly the same.
Re: Ask HN: Do you still use search engines?
#320I still use search engine. LLMs are great but often get outdated by 6-12 months. Usually, I search for random coding topics and asking LLMs will often reproduce the outdated top(most voted) answer verbatim. However, there are some nice folks at SO/SE who come back later and update the answers or submit a new one, but LLMs often don’t return these but continue producing various spin off of the top answer by modifying variable names or adding/removing comments.
Hence, search still remains my hope until SO and the likes decay.
Additionally, many search engines now already generate quick summaries or result snippets without a lot of prompt-fu, hence LLMs have actually become 40:60(llm:search) ratio day to day.