4 ms·

The point of the OP is that they claim they understand everything. Example: https://www.blog.google/products/search/search-language-understanding-bert/ https://

The point of the OP is that they claim they understand everything. Example: https://www.blog.google/products/search/search-language-understanding-bert/ https://www.blog.google/products/search/search-language-unde...

mherdeg 6y ago
OK, but what if an Amazon algorithm has actually learned that people who search for "shirt without stripes" are more likely to buy more things if the first image they see is a picture of a striped shirt?
- DenisM 6y ago
  It's a tricky balance between sales and relevance. If you don't watch for relevance you will end up showing only booze and porn in the commerce search results, border-line-porn in video search (true story), and so on. edit: fixed typo born -> porn
  - viridian 6y ago
    Could you explain the border-line-born bit? I don't get it.
    - giggles_giggles 6y ago
      It's a typo, he meant porn
- AmazingTurtle 6y ago
  OK, what if not? See, your comment makes no sense
- cscurmudgeon 6y ago
  Isn't this just a version of "God works in mysterious ways" applied to AI? "The AI works in mysterious ways. Trust it."
  - bradleybuda 6y ago
    It's the efficient markets hypothesis applied to AI. "If the AI could make more money by showing something else, it would be showing something else".
    - ghostwriter 6y ago
      in that case the applied AI stops being a Search tool (as was the purpose of the search bar) and becomes a new Ad tool. And this masquerading is not a great thing at all, for the same reason why people don't like bots pretending being humans during phone calls.
      - etherealG 6y ago
        But it’s not a search tool. It’s a make Amazon money tool, as all their tools are. I think you misunderstand why amazon build these tools if you think they are to make your life easier in trying to locate things to buy on their site. That’s a happy coincidence. They build them to make money.
    - Stranger43 6y ago
      The original version of which talks of an "Invisible hand" and if that's not an metaphor for divine intervention hidden by an chaotic system i don't know what is. It's rather surprising how often almost all complex systems theories be it AI, cosmology or economics have an aspects where even the theorists are resorting "to because it is". Sometimes those statements are based on measured data but it's not always easy or possible to do so accurately for highly interconnected system or worse system where you have actors reacting to theoretical model in a way that changes how the system behaves.
    - bscphil 6y ago
      I don't think it violates the efficient markets hypothesis. I think that entails "if it was possible for someone to create an AI that would make more money by showing something else, the AI would be showing something else." Roughly, efficiency means that the most profitable solution will be the one that wins out, but "profitability" includes the costs of building such an advanced system, assuming that it's possible at all.
- rhizome 6y ago
  Then the algorithm would not be acting in the customer's interests. Presumably this would be after the algo devalued people who clicked on "Next Page" until they came to a page that had stripeless shirts on it, or who, after the search, only ever clicked on stripeless shirts. "Deeds not words," dontchaknow.
  - etherealG 6y ago
    But the algorithm wasn’t programmed to act in the customer’s interest. It was programmed in amazons.
- parsimo2010 6y ago
  I think it's more likely that they've A/B tested faulty algorithms and picked the one that's the least faulty. I don't have proof, but I strongly believe that a search algorithm that returns what a customer is actually searching for will drive more sales. I suppose it's possible that with time, consistently bad results will beat a customer into submission and drive more sales of stuff the customer doesn't want. But I don't believe that's true, and this would only be the case if the customer accepts that the thing they want doesn't exist. If the customer is pretty sure that solid color shirts exist, they'll just shop elsewhere until they find it.
- yann63 6y ago
  If that's true, should we have the same kind of results when searching with "plain shirts", then? Which is not the case: searching for "plain shirts" do not give similar results than searching for "shirts without stripes".
jkaptur 6y ago
From that link: "So that’s a lot of technical details, but what does it all mean for you? Well, by applying BERT models to both ranking and featured snippets in Search, we’re able to do a much better job helping you find useful information. In fact, when it comes to ranking results, BERT will help Search better understand one in 10 searches in the U.S. in English, and we’ll bring this to more languages and locales over time. "Particularly for longer, more conversational queries, or searches where prepositions like “for” and “to” matter a lot to the meaning, Search will be able to understand the context of the words in your query. You can search in a way that feels natural for you. ... "No matter what you’re looking for, or what language you speak, we hope you’re able to let go of some of your keyword-ese and search in a way that feels natural for you. But you’ll still stump Google from time to time. Even with BERT, we don’t always get it right. If you search for “what state is south of Nebraska,” BERT’s best guess is a community called “South Nebraska.” (If you've got a feeling it's not in Kansas, you're right.) "Language understanding remains an ongoing challenge, and it keeps us motivated to continue to improve Search. We’re always getting better and working to find the meaning in-- and most helpful information for-- every query you send our way."
TulliusCicero 6y ago
Just because they've implemented something that helps in a certain area doesn't mean they'll always get it right. I think you'd struggle to find anywhere Google claims to "understand everything", making your assertion a strawman. Literally in the article you're quoting from Google: > But you’ll still stump Google from time to time. Even with BERT, we don’t always get it right. If you search for “what state is south of Nebraska,” BERT’s best guess is a community called “South Nebraska.” (If you've got a feeling it's not in Kansas, you're right.)
lonelappde 6y ago
Why do people make false claims while linking directly to a primary sources that unequivocally and repeatedly contradicts their claim? > sometimes still don’t quite get it right, > Even with BERT, we don’t always get it right. And nothing in the blog is about image search.
- onurcel 6y ago
  The fact they put some disclaimer after bragging doesn't discredit the point of the author who says those companies brag with AI but still fail to fulfil basic queries.

mherdeg 6y ago

OK, but what if an Amazon algorithm has actually learned that people who search for "shirt without stripes" are more likely to buy more things if the first image they see is a picture of a striped shirt?

DenisM 6y ago
It's a tricky balance between sales and relevance. If you don't watch for relevance you will end up showing only booze and porn in the commerce search results, border-line-porn in video search (true story), and so on. edit: fixed typo born -> porn
- viridian 6y ago
  Could you explain the border-line-born bit? I don't get it.
  - giggles_giggles 6y ago
    It's a typo, he meant porn
AmazingTurtle 6y ago
OK, what if not? See, your comment makes no sense
cscurmudgeon 6y ago
Isn't this just a version of "God works in mysterious ways" applied to AI? "The AI works in mysterious ways. Trust it."
- bradleybuda 6y ago
  It's the efficient markets hypothesis applied to AI. "If the AI could make more money by showing something else, it would be showing something else".
  - ghostwriter 6y ago
    in that case the applied AI stops being a Search tool (as was the purpose of the search bar) and becomes a new Ad tool. And this masquerading is not a great thing at all, for the same reason why people don't like bots pretending being humans during phone calls.
    - etherealG 6y ago
      But it’s not a search tool. It’s a make Amazon money tool, as all their tools are. I think you misunderstand why amazon build these tools if you think they are to make your life easier in trying to locate things to buy on their site. That’s a happy coincidence. They build them to make money.
  - Stranger43 6y ago
    The original version of which talks of an "Invisible hand" and if that's not an metaphor for divine intervention hidden by an chaotic system i don't know what is. It's rather surprising how often almost all complex systems theories be it AI, cosmology or economics have an aspects where even the theorists are resorting "to because it is". Sometimes those statements are based on measured data but it's not always easy or possible to do so accurately for highly interconnected system or worse system where you have actors reacting to theoretical model in a way that changes how the system behaves.
  - bscphil 6y ago
    I don't think it violates the efficient markets hypothesis. I think that entails "if it was possible for someone to create an AI that would make more money by showing something else, the AI would be showing something else." Roughly, efficiency means that the most profitable solution will be the one that wins out, but "profitability" includes the costs of building such an advanced system, assuming that it's possible at all.
rhizome 6y ago
Then the algorithm would not be acting in the customer's interests. Presumably this would be after the algo devalued people who clicked on "Next Page" until they came to a page that had stripeless shirts on it, or who, after the search, only ever clicked on stripeless shirts. "Deeds not words," dontchaknow.
- etherealG 6y ago
  But the algorithm wasn’t programmed to act in the customer’s interest. It was programmed in amazons.
parsimo2010 6y ago
I think it's more likely that they've A/B tested faulty algorithms and picked the one that's the least faulty. I don't have proof, but I strongly believe that a search algorithm that returns what a customer is actually searching for will drive more sales. I suppose it's possible that with time, consistently bad results will beat a customer into submission and drive more sales of stuff the customer doesn't want. But I don't believe that's true, and this would only be the case if the customer accepts that the thing they want doesn't exist. If the customer is pretty sure that solid color shirts exist, they'll just shop elsewhere until they find it.
yann63 6y ago
If that's true, should we have the same kind of results when searching with "plain shirts", then? Which is not the case: searching for "plain shirts" do not give similar results than searching for "shirts without stripes".

jkaptur 6y ago

From that link: "So that’s a lot of technical details, but what does it all mean for you? Well, by applying BERT models to both ranking and featured snippets in Search, we’re able to do a much better job helping you find useful information. In fact, when it comes to ranking results, BERT will help Search better understand one in 10 searches in the U.S. in English, and we’ll bring this to more languages and locales over time. "Particularly for longer, more conversational queries, or searches where prepositions like “for” and “to” matter a lot to the meaning, Search will be able to understand the context of the words in your query. You can search in a way that feels natural for you. ... "No matter what you’re looking for, or what language you speak, we hope you’re able to let go of some of your keyword-ese and search in a way that feels natural for you. But you’ll still stump Google from time to time. Even with BERT, we don’t always get it right. If you search for “what state is south of Nebraska,” BERT’s best guess is a community called “South Nebraska.” (If you've got a feeling it's not in Kansas, you're right.) "Language understanding remains an ongoing challenge, and it keeps us motivated to continue to improve Search. We’re always getting better and working to find the meaning in-- and most helpful information for-- every query you send our way."

TulliusCicero 6y ago

Just because they've implemented something that helps in a certain area doesn't mean they'll always get it right. I think you'd struggle to find anywhere Google claims to "understand everything", making your assertion a strawman. Literally in the article you're quoting from Google: > But you’ll still stump Google from time to time. Even with BERT, we don’t always get it right. If you search for “what state is south of Nebraska,” BERT’s best guess is a community called “South Nebraska.” (If you've got a feeling it's not in Kansas, you're right.)

lonelappde 6y ago

Why do people make false claims while linking directly to a primary sources that unequivocally and repeatedly contradicts their claim? > sometimes still don’t quite get it right, > Even with BERT, we don’t always get it right. And nothing in the blog is about image search.

onurcel 6y ago
The fact they put some disclaimer after bragging doesn't discredit the point of the author who says those companies brag with AI but still fail to fulfil basic queries.