8 ms·

Where do you start with this one? But okay, here goes: 1. The post assumes that the search result quality is not good. This simply is not the case. Users say t

Where do you start with this one? But okay, here goes: 1. The post assumes that the search result quality is not good. This simply is not the case. Users say the search results are good. Could they be better? Sure. Are they terrible? No. 2. The post assumes users want transparency on the search algorithm. This clearly is not the case. The whole reason that google became what it is today is because of the lack of clutter and good search results. If you remember Altavista then you remember users constantly having to fidget with all these knobs in the advanced section to get relevant search results. Lycos, excite, etc weren't much better. I remember using Google for the first time and it just worked. 3. The post assumes that users want deep customization. This again is not the case. Users want less clutter, less knobs, and a faster search experience. Google is delivering that. 4. The post assumes that any startup with a better algorithm can beat google. However this is not the case. Google has deep search integration with mobile providers such as it's own Android OS and also iOS. it also has it's own browser (Chrome) and has paid deals with Mozilla. Obviously, both browsers default to Google when you type something in the search field. Not to mention, the leading IOT devices all connect to Google to find results when it needs to search the web. And for safe measure, Google has come out with its own voice assistant device, Google home, to safe guard against any shenanigans from Apple's Siri and Amazon's Alexa that might happen in the future. It is heavily subsidizing the price to make sure it has a strong horse in the voice game. Integration aside, it also has scale. It crawls most of the internet several times a day. That's no small feat and it is quite hard for a startup to just replicate that with funding. I mean, even that won't cut it. Google is now even using machine learning to understand relevant pieces of information on a webpage so when you query it by typing on your keyboard, by voice, or future mind control it delivers the answer you were looking for. The only valid point the post makes is that Google has squashed competition by buying any startup that remotely could threaten it's position...for safe measure. I'll give the author that.

hackinthebochs 9y ago
I'll come out and say it: Google's search quality is terrible. It's better than the competition, but that's a low bar. Most people do pretty good with Google simply because 1. their searches are "solved" by simply reducing the search space to pages that contain the combination of rare words in the query (e.g. "correct battery horse staple" is referring to the xkcd comic with high probability), along with the social signal of backlink ranking, 2. Google has trained us to use it effectively and so our expectations for search match that which Google can effectively deliver, and 3. All the meta-signalling that Google uses to turn obtuse queries like "that movie with the guy who does the thing" into the correct answer. But in general, if you can't figure out the exact words that are on the page you're interested in, the illusion fails and it reveals itself to be barely more than grep on steroids. Search is still waiting to be blown wide open and it very well could be by a competitor.
- nradov 9y ago
  Sure but to actually accomplish that would require real AGI, or something very close to it.
Nuzzerino 9y ago
Hi, Definitely good points, but I disagree with them. 1) The issue is not with search result quality for a given query. In fact, I agree that such quality would be difficult to match. But this is a very specific, very polished feature. It is by no means the only way to deliver a quality user experience. Rather, the issue (and perhaps this is just one issue with many more I am unaware of) is that the search space of possible queries is limited to a small number of dimensions. 2) Definitely under no illusion that there is a demand for transparency on the algorithm. However, if that information is a successfully-guarded secret, nobody else is going to create an incremental improvement over that specific type of feature. 3) You have two incorrect assumptions here. Not all users want the same thing. But more importantly, there is absolutely no reason why more customization means that the UI must become more complex. A company like Google is smart enough to see that. If you have more customization, it can be offered through an API. Even if not offered through an API, you could see open-source browser plugins to create a custom page, with no clutter and no knobs. And likely a fast search experience with some effort. 4) Absolutely false! The point of my post was to illustrate that the failed strategy of "imitate the UI with a better/different algorithm under the hood" will never work. Your point regarding scale is a good one, and it is exactly part of the reason why I believe that peer to peer blockchain technologies will be the likely place for this dethroning to occur. Users can collectively contribute their own little piece of the network, sometimes even being paid by doing so (refer to Steemit for a proof-of-concept). There are a lot of interesting ideas in some of the Ethereum distributed app whitepapers. Most of those ideas will never be successfully built, but there are enough of them where we will see some successes. Then the industry will learn from the mistakes that were made, and we will see a new paradigm. I think this will happen rapidly over the course of a few years. I already have specific conceptual ideas for how a search engine might operate (URLs can be associated with a reputation rating, various tags, and so on, and not bound to a centralized server).
- electic 9y ago
  Ah, you wrote the quora post. 1. How else would you like the results to be delivered? It is delivered in browser's search bars. It is delivered by visiting Google.com. It is delivered by voice. It is delivered by mobile devices. It is delivered inside my Nintendo Wii, XBOX, PS3/4, and my Nintendo Switch. It's delivered on my crappy Vizio TV via smart apps. It's even delivered in my car. Google search results are everywhere. 2. I think that is kind of the point. However, you assuming Google is being evil by keeping it a secret. That's not fully the case. A lot of SEO companies would love to see how things are ranked so they can beat it on behalf of their clients. 3. They already have an API and if you really want to customize the query, you can add all kind of query modifiers in the UI. 4. As a developer myself in the crypto space, blockchains make a poor curated database for the very reasons why Google keeps things a secret. Anyone with money can pump the reputation of an entry on the blockchain to get better rankings thus lowering the quality of results. Second, most blockchains can't handle the query load. Three, dapps cost money to run. I can see a market for users submitting entries. If there are users using it, it might catch on in the submission department. However, this brings me to my next point. Four, the chain would not have all the websites on the net. Kind of expensive to get it all on there and not scalable. Lastly, someone has to pay to run the query side too. Would users pay to run a search query and get gamed incomplete results? No.
QAPereo 9y ago
Just to 1... do they know any better? What other experiences do they have to compare against?
tluyben2 9y ago
1. Google has become less intuitive to use, but given the right queries it will return the best answers on the first page which the others will not. Almost always the first result under the ads is what I was looking for and that has improved. My searches are programming and travel and both are spot on, but only when formulating things in english-logic speak and using the right nomenclature. If not, you still get related topics and usually still better than the competition, but meaningless in the intent of the search. If I had to use ddg or bing or such for work, I would lose a great amount of time going through pages; I cannot remember when I went to page 2 on Google. For programming questions, you can be sure if it is not on page 1, it was not answered. Page 2 and further will contain scraped SO results for instance (I would like Google to be more firm on that but he).
- Nuzzerino 9y ago
  There is an insane amount of content on the Internet which will not be returned in a query, no matter what you enter. More than half. More than 90%. Probably even close to 99%. Yes, you will get better results than the competition. No, it is not better by much. And I don't go to page 2 anymore because Page 2 will always have garbage on it. I remember in the old days, in the 90's, I could search and get 5 or 10 pages of unique pages. They might not have all been entirely relevant, but they might have been interesting reads nonetheless. With today's Google search, I get about 3 or 4 typical relevant results, often from the same websites vomited out from the top 50 from Alexa 500. And then a lot of spam, computer-generated web pages after that. There is no hope for anyone who wants to independently publish "a web page", anymore. I realize that this isn't something easily solved due to the fact that the Internet is vastly more populated now. But I don't think the system that we have in place, with Google, is anywhere near the level of sophistication that we can achieve as we approach the 2020s. We have this narrative that Google is such a vastly powerful collective, and that this is the best that they can come up with, and therefore that's what we have to work with, as Internet users. My post was not to show that I have all of the answers, but to illustrate the fallacious reasoning which deters people from even attempting to solve the problem. Folks, as long as we continue that mindset, that's how things are going to continue to be. /rant
  - 9y ago
RandomInteger4 9y ago
(1) My impression from using Google is that it's really great for technical / scientific first page search results, but pretty horrible otherwise. (3) No, I'm pretty sure users want more knobs or at least filter options. Hell, even a simple domain blacklist feature would be amazing so that I could get the woowoo psuedo science cruft -- which has been SEO'd to the front page -- out of my results, so I didn't have to put any mental effort into visually filtering out multiples from the same site. Hell, they could even collate results from the same domain into a single grid folder to optimize the time I spend searching through the search results. Don't conflate Google's lack of innovation for an optimized experience simply because it's Google. I love Google, but there is a massive amount of improvement that can go into their search user interface. Same for every other search engine, but Google is the leader in this space, so they get the brunt of the criticism, and should they change, then others can follow suit.
- tajen 9y ago
  > Hell, even a simple domain blacklist feature would be amazing so that I could get the woowoo psuedo science cruft -- which has been SEO'd to the front page -- out of my results You can put w3schools in /etc/hosts to an IP like 127.0.0.1. You still have to visually parse the search results, but if you accidentally click, you’re not giving a pageview by mistake.
- stevenhuang 9y ago
  > domain blacklist feature I use this Google plugin just for this: https://chrome.google.com/webstore/detail/personal-blocklist-by-goo/nolijncfnkgaikbjbdaogikpmpbdcdef https://chrome.google.com/webstore/detail/personal-blocklist...
  - RandomInteger4 9y ago
    Thanks for the heads up on this. I'll give it a spin.
- TulliusCicero 9y ago
  > No, I'm pretty sure users want more knobs or at least filter options. Nah, 90% either don't or wouldn't use them if they were there. Sure, if you ask people directly if they want more options, they'll say yes, who doesn't want more choices? But that's a different thing from actually using them when they're available. The vast majority of users for a utility service like a search engine just go with the defaults.
fpoling 9y ago
I use DuckDuckGo as my primary search engine. It is good enough. But their algorithm is different. It took me some time to adjust for that so my searches works like they did with Google. Which point out that at this point not only Google has a good search algorithm, but users are trained to help it do better. A search startup has to overcome that as well.
- piaste 9y ago
  Can you give some tips about how you adjusted your search queries? Whenever I tried to use DDG I wouldn't be able to get results from it that were as good as Google's.
  - ctrl-j 9y ago
    Indeed. I like the idea of DDG. However, I always ended up getting my results by doing "g! <my search query>" I stopped using it because it ended up being extra time spent running my google queries.
  - fpoling 9y ago
    This is very subjective and I am not sure that it even matches the reality (it could be that within the last 6 months the search engines changed), but with DuckDuckGo I stopped to use verbs at all. Another thing is that at least for searches about IT Google prioritise recently posted/changed pages more so with DuckDuckGo I often added 2017 to get the latest hype (soon to be changed to 2018 :) ). Whatever it was at the beginning of the switch I used to use Google on a daily basis when DDG results were not good. Then in few months it pretty much stopped.
- sjy 9y ago
  I remember being a little creeped out when Google first started automatically localising queries, but when I tried switching to DDG earlier this year I realised that I've become utterly dependent on it. I constantly had to think about whether to append my city or state to the query and even after doing so would often get American results. (For example, searching for anything relating to the WA government still turns up information about Washington, even when the country is set to Australia.)
d0mine 9y ago
I don't know why but google results became mediocre/bad recently. There were a time when I could pick [good] google results in a blind test (for programming topics) between several search engines (Bing, Yandex). Now, DuckDuckGo can return noticeably better results for some queries than Google.
- OJFord 9y ago
  DDG is also easy to use as a shortcut - `!python import` for example to go straight to Python docs search results. With Google without ! I'd be hunting for the result from that site, because site:... is longer to type, and I probably wouldn't remember the domain anyway.
hyperpallium 9y ago
5. Plain old capital-intensive advantage: local server farms, enabling low-latency search (and "suggest") results, which been found to be very important for users. One can interpret this (and android) as google desparately spending money whereever money can give some competitive advantage, no matter how small, because google is acutely aware of how easily users can switch search engines. It's not a super clever strategy, but just: what else can you do with all this cash?
proginthebox 9y ago
I agree with almost all the points apart from transparency. Transparency is not customization. Transparency is basically like, you search something google returns something, and then there is a link which tells you why google returned what it did. This is the transparency which is currently being pursued by the EU. That algorithm providers should be able to explain the results of their algorithm.
Eridrus 9y ago
IMO Google is unbeatable as long as nothing changes, if you really have to recreate Google as it is today, it will cost too much and you will never make it. But if there are technological shifts that invalidate the investment Google has made, then there is a chance. I think you're spot on about why Google is investing in the Assistant, but I think you underestimate the threat from Alexa. If Alexa gets a foot hold and manages to use the fact that you're having a dialog to deliver commercial queries better, Google is in a lot of trouble. Vertical-specific commerce sites are also a threat, if you no longer search Google for insurance, but go directly to an insurance aggregator, Google's search tech is irrelevant to capturing that advertising revenue. I think Craigslist is undergoing this problem since they've basically refused to invest in better UX, vertical specific sites are eating their lunch.
- jjeaff 9y ago
  I just don't see Alexa making huge inroads for shipping. It will be/is great for recurring purchases of stuff you already buy on Amazon, but people were already going directly to Amazon for that. But using a voice system to buy new stuff just isn't a good experience. It's like the old days of catalog shopping with a phone except there's no catalog. People need to pull up and see, look at reviews, watch YouTube review videos. Audio only just doesn't do that and is a step backwards. I don't even use Alexa for reorders because I want to make sure I'm still getting a good deal everytime I order. And for those that don't care about the deal, they can just setup subscription purchasing.
  - sjg007 9y ago
    Dunno... "Alexa I need toilet paper", "Alexa I need avocado toast".. pretty simple..
  - Eridrus 9y ago
    Some people need that, sure. I think Amazon may have a trust problem, but I've brought products based entirely on the Wirecutter/sweethome reviews, and I can easily imagine talking to their bot about what microwave they recommend. None of this works well now, but if Amazon builds a hardware moat, Google may not have a chance to compete with better software.

hackinthebochs 9y ago

I'll come out and say it: Google's search quality is terrible. It's better than the competition, but that's a low bar. Most people do pretty good with Google simply because 1. their searches are "solved" by simply reducing the search space to pages that contain the combination of rare words in the query (e.g. "correct battery horse staple" is referring to the xkcd comic with high probability), along with the social signal of backlink ranking, 2. Google has trained us to use it effectively and so our expectations for search match that which Google can effectively deliver, and 3. All the meta-signalling that Google uses to turn obtuse queries like "that movie with the guy who does the thing" into the correct answer. But in general, if you can't figure out the exact words that are on the page you're interested in, the illusion fails and it reveals itself to be barely more than grep on steroids. Search is still waiting to be blown wide open and it very well could be by a competitor.

nradov 9y ago
Sure but to actually accomplish that would require real AGI, or something very close to it.

Nuzzerino 9y ago

Hi, Definitely good points, but I disagree with them. 1) The issue is not with search result quality for a given query. In fact, I agree that such quality would be difficult to match. But this is a very specific, very polished feature. It is by no means the only way to deliver a quality user experience. Rather, the issue (and perhaps this is just one issue with many more I am unaware of) is that the search space of possible queries is limited to a small number of dimensions. 2) Definitely under no illusion that there is a demand for transparency on the algorithm. However, if that information is a successfully-guarded secret, nobody else is going to create an incremental improvement over that specific type of feature. 3) You have two incorrect assumptions here. Not all users want the same thing. But more importantly, there is absolutely no reason why more customization means that the UI must become more complex. A company like Google is smart enough to see that. If you have more customization, it can be offered through an API. Even if not offered through an API, you could see open-source browser plugins to create a custom page, with no clutter and no knobs. And likely a fast search experience with some effort. 4) Absolutely false! The point of my post was to illustrate that the failed strategy of "imitate the UI with a better/different algorithm under the hood" will never work. Your point regarding scale is a good one, and it is exactly part of the reason why I believe that peer to peer blockchain technologies will be the likely place for this dethroning to occur. Users can collectively contribute their own little piece of the network, sometimes even being paid by doing so (refer to Steemit for a proof-of-concept). There are a lot of interesting ideas in some of the Ethereum distributed app whitepapers. Most of those ideas will never be successfully built, but there are enough of them where we will see some successes. Then the industry will learn from the mistakes that were made, and we will see a new paradigm. I think this will happen rapidly over the course of a few years. I already have specific conceptual ideas for how a search engine might operate (URLs can be associated with a reputation rating, various tags, and so on, and not bound to a centralized server).

electic 9y ago
Ah, you wrote the quora post. 1. How else would you like the results to be delivered? It is delivered in browser's search bars. It is delivered by visiting Google.com. It is delivered by voice. It is delivered by mobile devices. It is delivered inside my Nintendo Wii, XBOX, PS3/4, and my Nintendo Switch. It's delivered on my crappy Vizio TV via smart apps. It's even delivered in my car. Google search results are everywhere. 2. I think that is kind of the point. However, you assuming Google is being evil by keeping it a secret. That's not fully the case. A lot of SEO companies would love to see how things are ranked so they can beat it on behalf of their clients. 3. They already have an API and if you really want to customize the query, you can add all kind of query modifiers in the UI. 4. As a developer myself in the crypto space, blockchains make a poor curated database for the very reasons why Google keeps things a secret. Anyone with money can pump the reputation of an entry on the blockchain to get better rankings thus lowering the quality of results. Second, most blockchains can't handle the query load. Three, dapps cost money to run. I can see a market for users submitting entries. If there are users using it, it might catch on in the submission department. However, this brings me to my next point. Four, the chain would not have all the websites on the net. Kind of expensive to get it all on there and not scalable. Lastly, someone has to pay to run the query side too. Would users pay to run a search query and get gamed incomplete results? No.

QAPereo 9y ago

Just to 1... do they know any better? What other experiences do they have to compare against?

tluyben2 9y ago

1. Google has become less intuitive to use, but given the right queries it will return the best answers on the first page which the others will not. Almost always the first result under the ads is what I was looking for and that has improved. My searches are programming and travel and both are spot on, but only when formulating things in english-logic speak and using the right nomenclature. If not, you still get related topics and usually still better than the competition, but meaningless in the intent of the search. If I had to use ddg or bing or such for work, I would lose a great amount of time going through pages; I cannot remember when I went to page 2 on Google. For programming questions, you can be sure if it is not on page 1, it was not answered. Page 2 and further will contain scraped SO results for instance (I would like Google to be more firm on that but he).

Nuzzerino 9y ago
There is an insane amount of content on the Internet which will not be returned in a query, no matter what you enter. More than half. More than 90%. Probably even close to 99%. Yes, you will get better results than the competition. No, it is not better by much. And I don't go to page 2 anymore because Page 2 will always have garbage on it. I remember in the old days, in the 90's, I could search and get 5 or 10 pages of unique pages. They might not have all been entirely relevant, but they might have been interesting reads nonetheless. With today's Google search, I get about 3 or 4 typical relevant results, often from the same websites vomited out from the top 50 from Alexa 500. And then a lot of spam, computer-generated web pages after that. There is no hope for anyone who wants to independently publish "a web page", anymore. I realize that this isn't something easily solved due to the fact that the Internet is vastly more populated now. But I don't think the system that we have in place, with Google, is anywhere near the level of sophistication that we can achieve as we approach the 2020s. We have this narrative that Google is such a vastly powerful collective, and that this is the best that they can come up with, and therefore that's what we have to work with, as Internet users. My post was not to show that I have all of the answers, but to illustrate the fallacious reasoning which deters people from even attempting to solve the problem. Folks, as long as we continue that mindset, that's how things are going to continue to be. /rant
- 9y ago

RandomInteger4 9y ago

(1) My impression from using Google is that it's really great for technical / scientific first page search results, but pretty horrible otherwise. (3) No, I'm pretty sure users want more knobs or at least filter options. Hell, even a simple domain blacklist feature would be amazing so that I could get the woowoo psuedo science cruft -- which has been SEO'd to the front page -- out of my results, so I didn't have to put any mental effort into visually filtering out multiples from the same site. Hell, they could even collate results from the same domain into a single grid folder to optimize the time I spend searching through the search results. Don't conflate Google's lack of innovation for an optimized experience simply because it's Google. I love Google, but there is a massive amount of improvement that can go into their search user interface. Same for every other search engine, but Google is the leader in this space, so they get the brunt of the criticism, and should they change, then others can follow suit.

tajen 9y ago
> Hell, even a simple domain blacklist feature would be amazing so that I could get the woowoo psuedo science cruft -- which has been SEO'd to the front page -- out of my results You can put w3schools in /etc/hosts to an IP like 127.0.0.1. You still have to visually parse the search results, but if you accidentally click, you’re not giving a pageview by mistake.
stevenhuang 9y ago
> domain blacklist feature I use this Google plugin just for this: https://chrome.google.com/webstore/detail/personal-blocklist-by-goo/nolijncfnkgaikbjbdaogikpmpbdcdef https://chrome.google.com/webstore/detail/personal-blocklist...
- RandomInteger4 9y ago
  Thanks for the heads up on this. I'll give it a spin.
TulliusCicero 9y ago
> No, I'm pretty sure users want more knobs or at least filter options. Nah, 90% either don't or wouldn't use them if they were there. Sure, if you ask people directly if they want more options, they'll say yes, who doesn't want more choices? But that's a different thing from actually using them when they're available. The vast majority of users for a utility service like a search engine just go with the defaults.

fpoling 9y ago

I use DuckDuckGo as my primary search engine. It is good enough. But their algorithm is different. It took me some time to adjust for that so my searches works like they did with Google. Which point out that at this point not only Google has a good search algorithm, but users are trained to help it do better. A search startup has to overcome that as well.

piaste 9y ago
Can you give some tips about how you adjusted your search queries? Whenever I tried to use DDG I wouldn't be able to get results from it that were as good as Google's.
- ctrl-j 9y ago
  Indeed. I like the idea of DDG. However, I always ended up getting my results by doing "g! <my search query>" I stopped using it because it ended up being extra time spent running my google queries.
- fpoling 9y ago
  This is very subjective and I am not sure that it even matches the reality (it could be that within the last 6 months the search engines changed), but with DuckDuckGo I stopped to use verbs at all. Another thing is that at least for searches about IT Google prioritise recently posted/changed pages more so with DuckDuckGo I often added 2017 to get the latest hype (soon to be changed to 2018 :) ). Whatever it was at the beginning of the switch I used to use Google on a daily basis when DDG results were not good. Then in few months it pretty much stopped.
sjy 9y ago
I remember being a little creeped out when Google first started automatically localising queries, but when I tried switching to DDG earlier this year I realised that I've become utterly dependent on it. I constantly had to think about whether to append my city or state to the query and even after doing so would often get American results. (For example, searching for anything relating to the WA government still turns up information about Washington, even when the country is set to Australia.)

d0mine 9y ago

I don't know why but google results became mediocre/bad recently. There were a time when I could pick [good] google results in a blind test (for programming topics) between several search engines (Bing, Yandex). Now, DuckDuckGo can return noticeably better results for some queries than Google.

OJFord 9y ago
DDG is also easy to use as a shortcut - `!python import` for example to go straight to Python docs search results. With Google without ! I'd be hunting for the result from that site, because site:... is longer to type, and I probably wouldn't remember the domain anyway.

hyperpallium 9y ago

5. Plain old capital-intensive advantage: local server farms, enabling low-latency search (and "suggest") results, which been found to be very important for users. One can interpret this (and android) as google desparately spending money whereever money can give some competitive advantage, no matter how small, because google is acutely aware of how easily users can switch search engines. It's not a super clever strategy, but just: what else can you do with all this cash?

proginthebox 9y ago

I agree with almost all the points apart from transparency. Transparency is not customization. Transparency is basically like, you search something google returns something, and then there is a link which tells you why google returned what it did. This is the transparency which is currently being pursued by the EU. That algorithm providers should be able to explain the results of their algorithm.

Eridrus 9y ago

IMO Google is unbeatable as long as nothing changes, if you really have to recreate Google as it is today, it will cost too much and you will never make it. But if there are technological shifts that invalidate the investment Google has made, then there is a chance. I think you're spot on about why Google is investing in the Assistant, but I think you underestimate the threat from Alexa. If Alexa gets a foot hold and manages to use the fact that you're having a dialog to deliver commercial queries better, Google is in a lot of trouble. Vertical-specific commerce sites are also a threat, if you no longer search Google for insurance, but go directly to an insurance aggregator, Google's search tech is irrelevant to capturing that advertising revenue. I think Craigslist is undergoing this problem since they've basically refused to invest in better UX, vertical specific sites are eating their lunch.

jjeaff 9y ago
I just don't see Alexa making huge inroads for shipping. It will be/is great for recurring purchases of stuff you already buy on Amazon, but people were already going directly to Amazon for that. But using a voice system to buy new stuff just isn't a good experience. It's like the old days of catalog shopping with a phone except there's no catalog. People need to pull up and see, look at reviews, watch YouTube review videos. Audio only just doesn't do that and is a step backwards. I don't even use Alexa for reorders because I want to make sure I'm still getting a good deal everytime I order. And for those that don't care about the deal, they can just setup subscription purchasing.
- sjg007 9y ago
  Dunno... "Alexa I need toilet paper", "Alexa I need avocado toast".. pretty simple..
- Eridrus 9y ago
  Some people need that, sure. I think Amazon may have a trust problem, but I've brought products based entirely on the Wirecutter/sweethome reviews, and I can easily imagine talking to their bot about what microwave they recommend. None of this works well now, but if Amazon builds a hardware moat, Google may not have a chance to compete with better software.