Avid Amoeba is right that Google ruined their own search before LLMs entered the public consciousness (this does not mean LLMs didn’t exist before this, but that they were not widely available for the general public to use or became part of the zeitgeist).
If you don’t agree please listen to the Better Offline podcast episode “The Man That Destroyed Google Search”. The episode goes through the rollbacks/changes Google made to their search Algorithm well before AI was commonplace.
Yeah. Also I’m guessing their AI additions to search made their profit margins worse since they take a lot more computation to produce. Although they probably cache a lot of them for common searches.
Even though that surely results in them being able to access more money and makes shareholders richer, that’s not a factor in profit margins. Profit margins are just about revenue vs cost. In this case - how much the make from each search vs how much it costs to produce that search.
Many people around me are using LLMs in many parts of their work al the time. Neutral networks are used in many useful situations. I feel exactly like you, but I’m afraid we’re going to have to cope with it.
The US National Weather Service releases updated 84-hour forecasts every 6 hours. Even with supercomputers at their disposal, due to the computational complexity of simulating physics, that is their best possible effort.
Google, meanwhile, is “developing a machine learning model that it says can accurately predict weather in seconds – not hours – and outperforms 90% of the targets used by the world’s best weather prediction systems.” Using a single desktop computer, they can generate a highly accurate 10-day forecast in under a minute.
Yes. Search generally pulls data from databases. It doesn’t compute weather forecasts. The addition of AI results is net addition computation. In the worst case scenario where the generation of the AI results happens on-the-fly, that’s a lot more computation. I’m sure they pre-compute a lot of them so they’re not in the worst case scenario. However in the best case scenario they still have to do this new additional heavy (check LLM compute usage) computation once per result. So the profit margin for search is very likely lower than it used to be when isolating for this variable. If they’re somehow increasing their revenue from these results, that’s another variable that might offset it. I’ve no idea. What I’m certain about is the cost is higher after AI results were introduced because more energy is used.
That depends strongly on which “public” we are talking about - some extremely intelligent people I have talked to don’t even know what Reddit is. Old Google searches got bad, but if you scrolled down far enough, or added “reddit” to the search terms, they used to be salvageable. So it’s less of a hard cutoff and more of a long process that brought us to where we are today.
Public in this term has nothing to do with intelligence, but rather people outside of companies working on AI/LLMs or doing AI research. It’s why I mentioned it entering the zeitgeist.
I never mentioned a hard cutoff but said they ruined it before LLMs were in use by the general public. Essentially I’m referring to the starting of the degradation of Google’s search which they made conscious decisions that deliberately put profit above all.
My apologies that me being hyperbolic did not add clarity and instead caused confusion:-). Ultimately I agree, but was adding the point that users who were either savvy or dedicated enough could still get a lot of use out of Google until more recently, whereupon it is now just a huge mess that makes it more worthwhile to abandon completely (in favor of e.g. DuckDuckGo) - even though it was the demise of Reddit rather than the addition of LLMs that caused the sharp decline (+ other things too, e.g. there was a strike of mods at StackOverflow), i.e. Reddit (& others) was propping up Google results for the longest time, which does not excuse Google for allowing such instability, but helps explain the timeline wherein Google results were both “usable” (even if less so than the past) and also “degraded” at the same time.
It’s all good, we both clarified our* thoughts on the matter and to be fair using “ruined” instead of “ruining” or “started to ruin” indicates a completed process or final state instead of a continuous one.
I agree that previously one could construct a search to sort the noise out, but as you stated this has become unfeasible without a sharp increase of queries needed to refine results which has shifted the thought from questioning if Google search is bad to now generally accepted belief - to the point where people are trying to quantify and provide evidence to back up the claim.
If I can go on a tangent: it is conversations like this that continually convince me that I need never go back to Reddit. Not EVERY SINGLE conversation needs to be full of snark and vitriol. Being able to discuss things rationally, calmly, and with kindness is possible, if only people will create the space within which they are allowed to happen:-).
And how that relates is: using DuckDuckGo convinced me similarly to abandon Google:-). Caveats include using Google Images, Google Maps, etc. e.g. to look up the hours of a shop (the SEO optimization there works for rather than against me, although tbf quite often I have to bat away unrelated results vying for my increased attention due merely to having paid for that exact privilege), but overall the results of DDG are just extremely much more worth my time than Google’s.
As an example, if you search for the keyword “Lemmy”, DDG pulls up Lemmy.World as the #2 hit (which notably has ~80% of all active users on Lemmy, so is overwhelmingly deserving of being listed so highly), after the #1 hit being the singer, whereas on Google the first instance mentioned is Lemmy.ml (that has 2,206 active monthly users, compared to Lemmy.World’s 17,122 that is roughly an order of magnitude higher, and also housing the most-used communities e.g. !technology@lemmy.world has 16.9k active monthly users compared to !asklemmy@lemmy.ml’s top community with 8.44K), and that not until the #4 hit.
i.e., not only are Google results commodified, but as you said they are “ruined” as well - to the point of representing actual & active disinformation (for the sake of $$$) rather than merely misinformation (aka oopsies). We can scroll past one, two, even ten ads, but how do we find our info when the sorting refuses to distinguish between SEO-advanced results and “real” ones? I dunno, perhaps the above one is a poor example (edit: b/c in the past, Lemmy.ml really was the top Lemmy instance, for so very long), but I think you know what I mean regardless:-).
They specifically made search less accurate so that users would search multiple times to boost the number of ads that get displayed to juice their numbers for quarterly earnings. You can blame Prabhakar Raghavan.
They’re even shoving AI into Youtube by placing a summary in plain text below some videos now. Don’t know if it’s opt-in or just randomly placed for testing but so far I’m not impressed because it skips over important things. I’m honestly puzzled as to why the hell they’re doing this.
I do like how AI works for referencing articles. You can tap on any sentence in the summary and it will display all links that contain that source information. It’s actually pretty useful.
I find that in many cases, if you actually click the link to find the sourced information, it’s not there. I’ve experienced this with nearly every LLM front-end platform.
Another problem is they ruined their own search with AI.
Kicked themselves right in the nuts.
They ruined it without AI before AI was commonplace. They ruined it with higher profit margins. 🥹
Avid Amoeba is right that Google ruined their own search before LLMs entered the public consciousness (this does not mean LLMs didn’t exist before this, but that they were not widely available for the general public to use or became part of the zeitgeist).
If you don’t agree please listen to the Better Offline podcast episode “The Man That Destroyed Google Search”. The episode goes through the rollbacks/changes Google made to their search Algorithm well before AI was commonplace.
Yeah. Also I’m guessing their AI additions to search made their profit margins worse since they take a lot more computation to produce. Although they probably cache a lot of them for common searches.
Probably made the margins better because investors apparently still love hearing the word “AI” attached to shit
Even though that surely results in them being able to access more money and makes shareholders richer, that’s not a factor in profit margins. Profit margins are just about revenue vs cost. In this case - how much the make from each search vs how much it costs to produce that search.
I hope AI is the new metaverse. I’ll have a good chuckle when it all implodes.
Many people around me are using LLMs in many parts of their work al the time. Neutral networks are used in many useful situations. I feel exactly like you, but I’m afraid we’re going to have to cope with it.
The US National Weather Service releases updated 84-hour forecasts every 6 hours. Even with supercomputers at their disposal, due to the computational complexity of simulating physics, that is their best possible effort.
Google, meanwhile, is “developing a machine learning model that it says can accurately predict weather in seconds – not hours – and outperforms 90% of the targets used by the world’s best weather prediction systems.” Using a single desktop computer, they can generate a highly accurate 10-day forecast in under a minute.
More information:
https://www.weforum.org/stories/2023/12/ai-weather-forecasting-climate-crisis/
Given this information, and given the enshittification of Google search, would you still make the same guess about their profit margins?
Yes. Search generally pulls data from databases. It doesn’t compute weather forecasts. The addition of AI results is net addition computation. In the worst case scenario where the generation of the AI results happens on-the-fly, that’s a lot more computation. I’m sure they pre-compute a lot of them so they’re not in the worst case scenario. However in the best case scenario they still have to do this new additional heavy (check LLM compute usage) computation once per result. So the profit margin for search is very likely lower than it used to be when isolating for this variable. If they’re somehow increasing their revenue from these results, that’s another variable that might offset it. I’ve no idea. What I’m certain about is the cost is higher after AI results were introduced because more energy is used.
That depends strongly on which “public” we are talking about - some extremely intelligent people I have talked to don’t even know what Reddit is. Old Google searches got bad, but if you scrolled down far enough, or added “reddit” to the search terms, they used to be salvageable. So it’s less of a hard cutoff and more of a long process that brought us to where we are today.
Public in this term has nothing to do with intelligence, but rather people outside of companies working on AI/LLMs or doing AI research. It’s why I mentioned it entering the zeitgeist.
I never mentioned a hard cutoff but said they ruined it before LLMs were in use by the general public. Essentially I’m referring to the starting of the degradation of Google’s search which they made conscious decisions that deliberately put profit above all.
My apologies that me being hyperbolic did not add clarity and instead caused confusion:-). Ultimately I agree, but was adding the point that users who were either savvy or dedicated enough could still get a lot of use out of Google until more recently, whereupon it is now just a huge mess that makes it more worthwhile to abandon completely (in favor of e.g. DuckDuckGo) - even though it was the demise of Reddit rather than the addition of LLMs that caused the sharp decline (+ other things too, e.g. there was a strike of mods at StackOverflow), i.e. Reddit (& others) was propping up Google results for the longest time, which does not excuse Google for allowing such instability, but helps explain the timeline wherein Google results were both “usable” (even if less so than the past) and also “degraded” at the same time.
It’s all good, we both clarified our* thoughts on the matter and to be fair using “ruined” instead of “ruining” or “started to ruin” indicates a completed process or final state instead of a continuous one.
I agree that previously one could construct a search to sort the noise out, but as you stated this has become unfeasible without a sharp increase of queries needed to refine results which has shifted the thought from questioning if Google search is bad to now generally accepted belief - to the point where people are trying to quantify and provide evidence to back up the claim.
This article links to a research paper on the topic: https://www.fastcompany.com/91012311/is-google-getting-worse-this-is-what-leading-computer-scientists-say
*Fixed typo of ‘out’ to ‘our’
If I can go on a tangent: it is conversations like this that continually convince me that I need never go back to Reddit. Not EVERY SINGLE conversation needs to be full of snark and vitriol. Being able to discuss things rationally, calmly, and with kindness is possible, if only people will create the space within which they are allowed to happen:-).
And how that relates is: using DuckDuckGo convinced me similarly to abandon Google:-). Caveats include using Google Images, Google Maps, etc. e.g. to look up the hours of a shop (the SEO optimization there works for rather than against me, although tbf quite often I have to bat away unrelated results vying for my increased attention due merely to having paid for that exact privilege), but overall the results of DDG are just extremely much more worth my time than Google’s.
As an example, if you search for the keyword “Lemmy”, DDG pulls up Lemmy.World as the #2 hit (which notably has ~80% of all active users on Lemmy, so is overwhelmingly deserving of being listed so highly), after the #1 hit being the singer, whereas on Google the first instance mentioned is Lemmy.ml (that has 2,206 active monthly users, compared to Lemmy.World’s 17,122 that is roughly an order of magnitude higher, and also housing the most-used communities e.g. !technology@lemmy.world has 16.9k active monthly users compared to !asklemmy@lemmy.ml’s top community with 8.44K), and that not until the #4 hit.
i.e., not only are Google results commodified, but as you said they are “ruined” as well - to the point of representing actual & active disinformation (for the sake of $$$) rather than merely misinformation (aka oopsies). We can scroll past one, two, even ten ads, but how do we find our info when the sorting refuses to distinguish between SEO-advanced results and “real” ones? I dunno, perhaps the above one is a poor example (edit: b/c in the past, Lemmy.ml really was the top Lemmy instance, for so very long), but I think you know what I mean regardless:-).
They ruined it by setting themselves as untouchable and wanting bigger profit margins than “richer than God” money.
They specifically made search less accurate so that users would search multiple times to boost the number of ads that get displayed to juice their numbers for quarterly earnings. You can blame Prabhakar Raghavan.
Their search was shit before AI. Unless you like pinterest and quora spam.
Their search algorithm was great.
It’s fucking awful with our without AI in 2024.
They’re even shoving AI into Youtube by placing a summary in plain text below some videos now. Don’t know if it’s opt-in or just randomly placed for testing but so far I’m not impressed because it skips over important things. I’m honestly puzzled as to why the hell they’re doing this.
They get revenue from the pre roll ad while you read the summary. Then they don’t have to pay the creator when you click away before watching.
And then they’ll stop having creators creating free content for their advertisement sceme to work. Genius!
I do like how AI works for referencing articles. You can tap on any sentence in the summary and it will display all links that contain that source information. It’s actually pretty useful.
I find that in many cases, if you actually click the link to find the sourced information, it’s not there. I’ve experienced this with nearly every LLM front-end platform.