Live data from Hacker News

Claude can now search the web

anthropic.com

311–320 of 758 posts

Re: Claude can now search the web

#311
post #102

Searching the web is a great feature in theory, but every implementation I've used so far looks at the top X hits and then interprets it to be the correct answer. When you're talking to an LLM about popular topics or common errors, the top results are often just blogspam or unresolved forum posts, so the you never get an answer to your problem. More of an indicator that web search is more unusable than ever, but inte…

Lol so true

Re: Claude can now search the web

#314
post #189

Earlier quoted context omitted.

I mean, for those of us who used it since way before the '20s, it's not really a sentiment - it's a fact. You used to be able to type in 3 words and whatever error message your stack trace was showing, and the first 3 links returned were very likely a definitive source to solving your problem.Written by a human, and believe my word for it - it was much better back then than the crap you get out of torturing whatever…

"and believe my word for it - it was much better back then than the crap you get out of torturing whatever your LLM of choice is" I was around as well and my memories do not confirm this. But google search definitely degraded a lot.

Yes and no. You used to find niche websites more easily, but I vividly remember the frustration with ExpertsExchange results (with answers that were all paywalled).

Re: Claude can now search the web

#315

Earlier quoted context omitted.

>looks at the top X hits and then interprets it to be the correct answer. LLMs are truly reaching human-like behavior then

The longer I've been in the workforce, the more I realize most humans actually kind of suck at their jobs. LLMs being more human like is the opposite of what I want.

Ever heard the saying “good help is hard to find.”? It’s not bullshit, it really is.

Re: Claude can now search the web

#317
post #102

Searching the web is a great feature in theory, but every implementation I've used so far looks at the top X hits and then interprets it to be the correct answer. When you're talking to an LLM about popular topics or common errors, the top results are often just blogspam or unresolved forum posts, so the you never get an answer to your problem. More of an indicator that web search is more unusable than ever, but inte…

Overall LLMs (that I've tested) don't know how to use a search engine, their queries are bad and naive, probably because the way to use a search engine isn't part of training data, it's just something that people learn to do by using them. Maybe Google has the data to make LLMs good at using search engines but would it serve their business?

Re: Claude can now search the web

#318

Earlier quoted context omitted.

The longer I've been in the workforce, the more I realize most humans actually kind of suck at their jobs. LLMs being more human like is the opposite of what I want.

That could very well be because the jobs are effectively useless. By no means does that mean the people are, nor is what the income allows them to do. But most jobs sure do seem pointless.

https://news.ycombinator.com/item?id=36373190

Re: Claude can now search the web

#319

Earlier quoted context omitted.

Google search is crap. It seems to be a sentiment among many HNers, but is it really that bad? I mostly use it for programming, so documentation/forums and it works out greatly. For some queries it even returns personal blogs (which people seem to bash google for not happening). Of course there are some queries that return purely AI blogspam, but reformatting the query with a bit more thought usually solves it. I won…

I watched one of my friends who says Google is useless use Google one day. If I were looking for a song, I would type in something like “song used at beginning of X movie indie rock” He would type in “X songs.” I basically find everything in Google in one search and it takes him several. I type in my thought straight whereas he seems to treat Google like a dumb keyword index.

Google used to be a "dumb keyword index" in the past. It worked better that way. You had some modicum of control over the matching process. For the past 10 years or so, Google turned more into "try to guess what a novice normie means", which removes user control (no more actually working verbatim search or logical operators...), and... well I failed to develop a mental model of how exactly it works. It's not a proper keyword search anymore, and it's not a proper DWIM system with true understanding of natural language like LLMs are. It's something... in between, inferior to both.

Actually, typing out "what a novice normie means" made me realize what is the probable reason Google turned out the way it is: optimizing for new users. However, a growing userbase means most users are new to Internet in general, and (with big enough growth) most queries are issued by people who are trying a search engine out for the first time, and have no clue how or why it works - and those queries are exactly the kind of queries Google is now good at, queries like example you provided.

Re: Claude can now search the web

#320

Earlier quoted context omitted.

It's kinda like online games. Most people who play a game are not too great at it, a large subset is pretty good, and then it's smaller and smaller groups as the ability increases. At the top you get the people who are true pros, they write the books, the guides, they solve the hardest problems, and everyone looks up to them. But spin the wheel and get a random SWE to do some work? It's not gonna be far off from an r…

https://danluu.com/p95-skill/ Continues to apply

Competitive games are complex. It's hard to be 95% percentile. There are so many mistakes one can make, even if each individual mistake is unlikely, it's likely that a mistake will be made. I participate in Dota 2, and literally everyone makes noticeable mistakes, even including tier 1 pro players and the top ranked pub players. I honestly find it amazing how good people are given how complex the domain is.

Now scale that up 10x, because reality is at least an order of magnitude more complex than a video game.

Post reply on HN