Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Seems to me that Google is turning on a lot of deep learning models in production which aren't production ready.

Search has gone to shit recently.

Image search is hot garbage for anything but images with well known people / places / things.

You can't reliably filter or sort news results by date.

YouTube is filtering pro LGBT content but allowing weird Eastern European pedo videos.

The play store is apparently moderated by bots and doesn't have rules or policies in place about giving devs time to respond.

Something has changed dramatically at the company in the last 5 years. Would be interesting to zero in on it.



This sounds like exactly what you'd expect from a company that is bleeding talent and morale.

Note: I do not have any information that that is what is happening in Google; I'm just saying what my guess would be if I were presented the situation as "these are the symptoms at Company X. What is the problem?"


And you're likely not far off.

It's a pretty well tested principle that massive familial wealth usually disappears by the 3rd or 4th generation (because the descendents regress to the norm in terms of ability) and it looks like we're seeing the same at Google.

They've got too little "real" work for the number of hands they need to keep busy, so unqualified people are going back and "improving" the work of the 1st and 2nd generation.

But because of the company's size, it can't hire high risk shoot from the hip revolutionaries (not that they'd want to work at a buttoned up Corp anyway), so it's staffed with safe employees who don't really resemble the cowboys of the 1st and 2nd generations, and hence can't improve the systems any further.

Happens to most companies beyond a certain size.


That doesn't line up with the fact that Google kills existing apps all the time, and look at this new Google communications platform.


Doesn't it line up exactly?

They release a new chat app every year and put the previous years app on a 5 year sunset.

So you get to put a bunch of developers on the roof every year.

It's Keynes "break things to put people to work fixing them" at its finest.


I searched for an error message yesterday and Search removed 18 of the 20 keywords (they are all scratched out as 'missing'). The only remaining keywords were for the platform I had the error on. Thousands of results that were all completely irrelevant. I was completely stunned how something I rely on could suddenly do that. What is the point?

Google's monopoly has gotten to the point where they do not care how their products perform, just so long as the money continues to rain in.


> (they are all scratched out as 'missing')

The 'missing' notes are incredibly condescending. Not only are the results irrelevant/incorrect, they also had to arrogantly point out exactly how they are choosing to ignore the user's request.


Amen to that. I got to where I am as a sysadmin by knowing how to filter out keywords to make my search specific. Then Google throw all that away and decide for me what I want to search for. No Google, you don't tell me what to do.


Animaniacs Episode 7, "Piano Rag":

    [The Warners try to hide form Dr, Scratchandsniff in
     a piano recital, where Tymannini is about to perform]
  
    TYMP: Franz Schubert intended the scherzo
          to reflect the struggle between intellect
          and the creative process.
    TYMP: However, Schubert was simply incapable
          of expressing such delicate nuance.
          But, thanks to *my* genius,
          I will perform this great work,
          *not* as the composer wrote it,
          but as he *intended*.
DWIM ("Do What I Mean")[1] has always been impossible; it's a fallacious belief that what a user intended is different from what they actually said, and that this intent can somehow be divined form a few words, typos, etc.

> as a sysadmin

It doesn't even require any advanced technical knowledge. Most people understand basic keyword searching. Over the last ~6-12 months, I've seen a wide variety of people complain about Google ignoring their search terms.

[1] https://news.ycombinator.com/item?id=15628014


Happens to me all the time now too. Even when I put the phrase in quotes it still omits words or says "search for X, do your really want to search for Y"

But maybe our use cases are the outliers, and the new search is what works for most people.


> But maybe our use cases are the outliers, and the new search is what works for most people.

I think it's a probability problem more often expressed in terms of medical tests / false positives. If a million people search for <word> but 1% of those make a typo and search for <wrod> instead, whereas a thousand people deliberately search for <wrod>, Google's observation of a search for <wrod> is 90% likely to be a typo.

However well-intentioned, this can undermine one of the core, original features of a good search engine like Google: deriving value from the "long tail" of esoteric (and often very technical) quests for information.


Sure, correcting typos, I wouldn't hold that against them. However, when they intentionally scratch out search terms for no apparent reason and show extremely vague matches, it's completely useless. I was searching for a specific error message that was causing my MacBook to freeze on login. Instead, the only two terms Google preserved were 'mac os'. First result was Apple's own page on the OS. Every result was entirely useless. Fine, if there's no actual result for this error message, tell me, but don't make me scan through a page full of irrelevant links and make me figure out myself why they're irrelevant.


The real problem here is that Google does this even when we use double quotes and the verbatim option :-/

I switched to DDG a while ago, -it is about as good as Google now, much thanks to Googles own effort ower the last decade to nerf its own search engine.

Since they are now about equal I use the one that doesn't track me and doesn't support Google.


I've been using DDG for about half a year now, most of the time it does its job, but just yesterday, there was a case where I had to run back to Google: Searching for the "Ship of Theseus" with the words "ship replaced part by part" on Google lead me straight to the Wiki article, whereas on DDG...

Not sure if they're using a statistical (DDG could never compete, too few searches performed) or an algorithmic (DDG could, in time, get this far too) pattern to achieve correct results.

EDIT: It's about the tenth result on DDG, the last one to show before having to click to load more results. Not bad, actually.


I've just changed my browser default to DDG. Think I'm gonna start detox'ing myself from Google.


Which was their killer feature. Yahoo!™ Was great if you knew what curated content you already wanted.

But if you wanted some esoteric, bizarrely specific thing indexed on a single page on a single site, you went with Google.

They're turning into Yahoo.


I just tried "wrod". Google asks if I meant "word", but it does perform a search for "wrod".


Wrod is probably common enough and has enough real results that it works.

"Sulliven", a mount in northwestern Scotland, gets "fixed" to Sullivan.

It does work when quoted though


Google maps, Wikipedia and other sources apparently spell it Suliven [1], and a Google search for that, without quotes, returns it. Searching for "sulliven" (in quotes) finds a few (possibly mis-spelled?) references to the mountain (and some other things, mostly people named Sulliven.)

[1] https://www.google.com/maps/place/Suilven/@58.120899,-5.4389...


this behavior is worse than that.

it basically says, "I won't search for that keyword, here are some results without it"


> Something has changed dramatically at the company in the last 5 years

Founders effectively retired. Google has a new CEO


Funny cause I have just started to notice that as well, BUT I am mainly a DDG user that resort to using Google search in case I cannot find something satisfactory on my preferred search engine.

Speaking about YouTube, I think it might be recommending through other videos that you might have watched/liked that are not related to either, but somehow other users have watched/liked as well. So instead of having a relation A-B, it is now a relation A-C <-> D-B, you and the other users have watched/liked C and D, therefore you might want to watch B.


Yeah YouTube's recommendations don't really strike me as being based on comparison of the content or themes of the videos, but based pretty heavily on what other users have watched.

To be fair, codifying video into clusterable features is a lot harder than audio. But when you have Google levels of money I'd expect a bit more progress.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: