Hacker News Reader

Comment by cmckn | Hacker News Reader

I don't think the "problem" here is bad search results -- I think the lack of clicks is thanks to search results becoming much richer. The "card" style answers that show up more and more often mean I don't have to actually visit the site where the information originated (which publishers despise, for good reason). Take a common search I made last year as an example, "covid king county" -- I almost never ended up visiting the county's coronavirus dashboard because Google had the graphs I wanted in the search results.

I think it's a popular opinion on HN that Google search sucks, but I just don't agree. I used DDG on all my devices for the better part of last year, and bailed when I noticed by g! usage ratcheting up.

View on HN

It sucks at searching for exact things, it would often subtly by essentially change the meaning of the search. Like confusing searches for the desired "weight of scooter" with the max "weight of rider" on the scooter. What a shame, it was even a shopping related search, where the money are. They could have led me to a sale.

When people are trying variations on a search it's a clear sign of failure. They should do something about it, like use a verbose natural language interface or select a different strategy for ranking or enable the exact in-depth mode. Apparently the NLP community can do natural language Q&A in papers, but Google can't do it in search.

Other pain points: searching in depth all results on a query, not just the top skim and remembering context between searches.

And Google Assistant itself is too poorly integrated and dumb. Where's the GPT-3 like language skill? So many TPUs what are they doing all day long? They have more text and images than any other entity on the internet, Google bot has been sucking it up for so long (not to mention links, keywords and clicks). It should show in the quality of their AI. It's a shame to have OpenAI steal their thunder like this.

Unless you are a part fo google search engineering this is a lot of assumptions about their tech. How do you know they don't use those techniques?

Undeniably they are the largest, most successful at search and they spend huge amounts of money on it.

And google is probably the biggest producer of ML research including? including a TON of NLP as it's a driver of their profit.

Yeah, the parent and most its children are focusing on something not even related to the article — this is about providing what the user is looking for without the need for more clicks.

While it’s an unpopular opinion on HN, there’s no denying that from a user perspective, that’s only a good thing.

Card results might be a good thing overall, but they certainly aren't only good. In many cases, adding the cards takes revenue away from the exact people who collected or created the content to make them possible in the first place. We'll never know how many websites shut down or never got created to begin with given Google's history of crushing the revenue from various sites on a whim.

I definitely read this like the two of you. If the information being shown by Google results is what I need, there isn't a need to click further. On top of that, many of us have been "trained" that products like ZScaler are going to block most sites and register a hit with InfoSec. I'm not going to click on bobsfunmainframefacts.com if Google scraped the needed info for me.

It can be simultaneously true that Google's gotten better at displaying results pages that fully satisfy the user without the need for a click, and worse at other kinds of queries. The latter case -- searches which I have to rephrase, or that cause me to give up on the search entirely -- adds to the total number of searches that do not lead to clicks.

Reply on news.ycombinator.com