Hacker News Reader

Comment by jahnu | Hacker News Reader

They mastered search. They mastered mapping.

Search has stagnated. Google has gotten scarily good at anticipating what I want to know, but conspicuously bad at answering hard questions. There's tons of room for improvement.

Similarly, Google Maps has added features that make it a better map, and it's a great place to find what time a store closes and what their phone number is. But it's been essentially the same service, with iterative improvements, for years. There's a world of possible uses for that data beyond what we imagine when we think of a map.

> Google has gotten scarily good at anticipating what I want to know, but conspicuously bad at answering hard questions.

Well, 98 percent of the market is not asking hard questions, so by dropping support, they are actually making their search product more tailored towards their market.

I suspect Google Search has not actually had quality regressions, but rather that your expectations have gone up over time without you realising it.

This. It used to be the first page of results was good enough. Now if it is not in the top 3-4, I refine my query.

Yes, I might have phrased that poorly. Google Search still finds relevant web pages just as well as it did 5 years ago, perhaps better. It certainly has leveraged the data it collects on me to suggest search terms in a way that's almost eerie.

But it's still a tool to find relevant pages, not a tool to find answers. It's made some progress with the semantic web and knowledge graph, but that still provides basic facts about people, places, and things. At its core, it's still a portal. It's still just a much better version of Excite. Search today feels like it did in 1997: a "solved problem" ripe for disruption. Google has the talent, technology, and data to produce something an order of magnitude better.

I agree, I think search is ripe for disruption, Google is the old guard in this respect, I think they reached a crossroads where improvements in their search experience did not translate to increased revenue, and they stuck with this 'good enough' solution which made a predictable amount of money. I still show many users the 'filter by date' drop down and it's the first time they have seen it, it still does not offer me any way to restrict my searches to categories, though it certainly seems to have 'knowledge' based data in reach when it shows me adverts. There is surely much more that can be done, most competitors seem to want to match reach rather than innovate on the UI, I'm sure there are lots of startups out there, so if anyone wants to point me in the right direction feel free :)

>Search has stagnated. Google has gotten...

Who downvoted this and why?

I think the biggest issue with google right now is they have so many simple answers available. If I google some even slightly in depth problem, I'm bombarded with quarter explanations from "popular" sites like howtogeek, cnet wikihow, etc, which as far as I remember have never helped me one bit.

They seem to rely too much on site popularity, and not enough about quality of result, and god help you if you try to google a higher-level math problem.

Some of the updates over the last year or so will put Google in a better position to increasingly put quality over popularity. There's still going to be a problem that people seem to trust Wikihow, etc...

Regarding the math problems: do you have an example of a query for which there's a good result on the internet, but Google doesn't return it?

Is anything truly "mastered"?

It's not "perfect", it's just "mastered"

What's the difference?

Probably "mastered" mean constant quality.

A grandmaster chess player has mastered chess, but he can't play perfectly because he's still a human.

Just yesterday I got a blank result page for a sensible query, seems to have hit some rarely visited bucket combined with a busy machine somewhere or the like, and the background request timed out. Of course a refresh fixed it, but software is never perfect.

Reply on news.ycombinator.com