Google Minus Google News
February 23, 2024 3:55 AM Subscribe
"The featured filters — Images, Videos, Maps, Flights, Shopping, Perspectives, etc. — change and reorder depending on the search term, but this was different. I wasn’t seeing the News tab as an option for search after search, even if I went looking in the 'All filters' drop-down menu. I tried with 'Julian Assange,' 'public subsidies for sports stadiums,' and 'Reckon layoffs.' None showed the News filter as an option. The next day, on a different computer, my News filter was (blessedly) back. But a few other users confirmed I was not alone." Last year Google cut jobs in its news division. Where is Google putting its resources these days? Exactly where you'd expect.
AI is an area where we should all be asking “why are you putting money into this?”, because it serves absolutely everything it touches badly.
posted by The River Ivel at 4:51 AM on February 23 [31 favorites]
posted by The River Ivel at 4:51 AM on February 23 [31 favorites]
News disappearing from Google. Twitter unusable. Reddit selling itself to the highest bidder. Facebook a graveyard of misinformation.
All at a time when we've never needed more awareness of what's going on in the rest of the world.
Cool cool cool cool.
posted by fight or flight at 5:08 AM on February 23 [63 favorites]
All at a time when we've never needed more awareness of what's going on in the rest of the world.
Cool cool cool cool.
posted by fight or flight at 5:08 AM on February 23 [63 favorites]
I thought they were putting all the money into their shopping division, given how messed up interacting with any shopping-type search results has gotten in the past few weeks.
posted by clawsoon at 5:13 AM on February 23 [6 favorites]
posted by clawsoon at 5:13 AM on February 23 [6 favorites]
Yes, the shopping-type search results are a horror. The other day I was looking for something (not shopping, can't remember the topic), and somehow Google's algorithm misidentified me as shopping for whatever it was. There simply was no option to switch to another tab or whatever that would actually allow me to search for what I needed. I went to a topical website and found what I wanted, but it was disconcerting.
posted by cupcakeninja at 5:19 AM on February 23 [8 favorites]
posted by cupcakeninja at 5:19 AM on February 23 [8 favorites]
News disappearing from Google. Twitter unusable. Reddit selling itself to the highest bidder. Facebook a graveyard of misinformation.
Hell, there have even been very quiet rumors floating about lately that even Gmail could potentially be heading to the gallows in the near/mid future. Of course, that may just be a scare tactic to soften the world into accepting Gmail moving to a subscription model.
posted by Thorzdad at 5:21 AM on February 23 [5 favorites]
Hell, there have even been very quiet rumors floating about lately that even Gmail could potentially be heading to the gallows in the near/mid future. Of course, that may just be a scare tactic to soften the world into accepting Gmail moving to a subscription model.
posted by Thorzdad at 5:21 AM on February 23 [5 favorites]
Oh! Oh! I used Bing AI recently to good effect! (can't believe I just wrote that...). I work at a nonprofit and I had to write a riveting email about another nonprofit that serves as a qualifications and quality-assurance service that indexes and references other nonprofits! (my non-scientific analysis of this email is that it was read by me, my boss and maybe 2 other people out of the 1500 who received it).
This company has a weird recent history, was recently swallowed by a company with a different name, and took the new name. But the qualifications "seal" that everyone knows and that we proudly display has a different name on it than the company that awards them (at least for now, until they change everything over). This was all confusing as hell to write about in a short, succinct manner, and I wasn't sure I even understood what the full story was.
So I typed in a basic query about it into whatever Bing uses for "AI" (I put quotes because AI is neither artificial nor intelligent), and it spat out about three sentences that made sense! I took those sentences and researched their basic premises and those sentences helped me sum up in very few words a rather complicated, boring and dumb business situation. Again: I made sure to look up each detail in what Bing spat at me. And it was correct!
So, score one for "AI"!
And yes. I used Bing. Because that's what pops up on our IT-mandated browsers, and I was in a hurry.
posted by SoberHighland at 5:32 AM on February 23 [6 favorites]
This company has a weird recent history, was recently swallowed by a company with a different name, and took the new name. But the qualifications "seal" that everyone knows and that we proudly display has a different name on it than the company that awards them (at least for now, until they change everything over). This was all confusing as hell to write about in a short, succinct manner, and I wasn't sure I even understood what the full story was.
So I typed in a basic query about it into whatever Bing uses for "AI" (I put quotes because AI is neither artificial nor intelligent), and it spat out about three sentences that made sense! I took those sentences and researched their basic premises and those sentences helped me sum up in very few words a rather complicated, boring and dumb business situation. Again: I made sure to look up each detail in what Bing spat at me. And it was correct!
So, score one for "AI"!
And yes. I used Bing. Because that's what pops up on our IT-mandated browsers, and I was in a hurry.
posted by SoberHighland at 5:32 AM on February 23 [6 favorites]
Really was not expecting Google of all companies to make even the tiniest foray into the open source AI world despite the We Have No Moat memo.
As expected from them in everything these days it’s the semi-shitty weaksauce version (7 billion parameters vs llama-2’s 70 billion from Facebook), but it’s a baby step in the right direction.
Of course the training data info is incredibly vague (6 trillion tokens, with an unspecified split across “web documents”, programming examples, and math), because all the major players are terrified of liability exposure from transparency in that regard.
One very positive standout is the model release notes bias metrics section, which have a far more detailed breakdown even than Mixtral-8x7B’s (current “latest and greatest” general purpose LLM in open source AI), with Gemma-7B narrowly beating Mixtral which narrowly beats Llama-2 70B in bias reduction. Fingers crossed this represents the new basic expectation for major releases.
That said, while it doesn’t quite give apples-to-apples capability metrics, in general they appear pretty dismal relative to the other two. I’m not seeing a reason to choose it over Mixtral, personally.
The release notes link has an inference chat widget if you want to try it, but login required with a disclaimer that they’re data harvesting everything (of course) on it, so: burner account and avoid personal info or just run it locally (sample setup code for that is provided).
Overall very meh with a notable exception in their bias reduction publication and results.
into whatever Bing uses for "AI"
Customized ChatGPT-4 with a pretty substantial ruleset prefixing any query, both of which are completely unsurprising since Microsoft is OpenAI’s Daddy Warbucks and they have a bad history with early AI forays going DUI Mel Gibson on them (link to WP Tay article).
posted by Ryvar at 5:56 AM on February 23 [7 favorites]
As expected from them in everything these days it’s the semi-shitty weaksauce version (7 billion parameters vs llama-2’s 70 billion from Facebook), but it’s a baby step in the right direction.
Of course the training data info is incredibly vague (6 trillion tokens, with an unspecified split across “web documents”, programming examples, and math), because all the major players are terrified of liability exposure from transparency in that regard.
One very positive standout is the model release notes bias metrics section, which have a far more detailed breakdown even than Mixtral-8x7B’s (current “latest and greatest” general purpose LLM in open source AI), with Gemma-7B narrowly beating Mixtral which narrowly beats Llama-2 70B in bias reduction. Fingers crossed this represents the new basic expectation for major releases.
That said, while it doesn’t quite give apples-to-apples capability metrics, in general they appear pretty dismal relative to the other two. I’m not seeing a reason to choose it over Mixtral, personally.
The release notes link has an inference chat widget if you want to try it, but login required with a disclaimer that they’re data harvesting everything (of course) on it, so: burner account and avoid personal info or just run it locally (sample setup code for that is provided).
Overall very meh with a notable exception in their bias reduction publication and results.
into whatever Bing uses for "AI"
Customized ChatGPT-4 with a pretty substantial ruleset prefixing any query, both of which are completely unsurprising since Microsoft is OpenAI’s Daddy Warbucks and they have a bad history with early AI forays going DUI Mel Gibson on them (link to WP Tay article).
posted by Ryvar at 5:56 AM on February 23 [7 favorites]
I never noticed this because I never go to the main Google home page looking for news. I search so much news (for both work-related and personal reasons) that I pull up the Google News site directly. Actually, I don't even do that. Firefox knows me so well that all I have to do is type "n" into the address bar and the page pops up.
But what I can say is that over the past 18 months or two years the search and filter function of Google News just died to the point where it was unusable. Now, I thought this might have been a retaliation for the Canadian government trying to make Google and Facebook cough up money for news, but I had no way easy way to test the results outside of Canada.
Eventually, it go so bad that I switched to using Yahoo's news search to do what I needed to do. Yes, Yahoo and no, I couldn't believe it myself that I'd fallen back on that ancient (by Internet standards) site, but it certainly worked better than broken Google News.
Since the government and Google struck a deal, I've been finding that Google News has been a bit more responsive, but it's not back to the state where it used to be.
I wasn't expecting Google to do any development of its news search service, I was (stupidly in retrospect) expecting the company to maintain and support what it had created--something that really shouldn't cost that much money. Don't tinker with things, don't improve things, check to see the power is on and that the servers are running and that should have basically been it, but of course it's never that simple.
posted by sardonyx at 6:37 AM on February 23 [11 favorites]
But what I can say is that over the past 18 months or two years the search and filter function of Google News just died to the point where it was unusable. Now, I thought this might have been a retaliation for the Canadian government trying to make Google and Facebook cough up money for news, but I had no way easy way to test the results outside of Canada.
Eventually, it go so bad that I switched to using Yahoo's news search to do what I needed to do. Yes, Yahoo and no, I couldn't believe it myself that I'd fallen back on that ancient (by Internet standards) site, but it certainly worked better than broken Google News.
Since the government and Google struck a deal, I've been finding that Google News has been a bit more responsive, but it's not back to the state where it used to be.
I wasn't expecting Google to do any development of its news search service, I was (stupidly in retrospect) expecting the company to maintain and support what it had created--something that really shouldn't cost that much money. Don't tinker with things, don't improve things, check to see the power is on and that the servers are running and that should have basically been it, but of course it's never that simple.
posted by sardonyx at 6:37 AM on February 23 [11 favorites]
On Ash Wednesday I was trying to find a Catholic church near my office so I could ideally pop in, get ashes, and get out. The closest only did ashes as part of a mass, and the mass times didn't line up with my daily agenda, so I was hoping to find an ongoing distribution.
Google maps wouldn't return any result for "Catholic church." Or "Church" in general. Or "Iglesia." Looks like it is working again this morning, but I don't know whether they dropped it because of potential threat or because so many people searched for it that morning that they blocked it (something I hadn't seen since Michael Jackson died).
I didn't get ashes. I had to rely on good services going away as my momento mori to kick off Lent.
posted by thecaddy at 6:39 AM on February 23 [7 favorites]
Google maps wouldn't return any result for "Catholic church." Or "Church" in general. Or "Iglesia." Looks like it is working again this morning, but I don't know whether they dropped it because of potential threat or because so many people searched for it that morning that they blocked it (something I hadn't seen since Michael Jackson died).
I didn't get ashes. I had to rely on good services going away as my momento mori to kick off Lent.
posted by thecaddy at 6:39 AM on February 23 [7 favorites]
Google may be shifting resources to AI, but I suspect this experiment may also be driven by the company’s continued fraught relationship with news publishers.
Both Google and Facebook have had unhappy relationships with publishers for years, getting worse as publishers have lobbied governments to force the tech companies to pay license fees for links. When it becomes too expensive or they want to throw their weight around, the tech companies have been willing to block news sources. Witness Google’s eight-year ban on Google News in Spain, and Facebook’s continuing (?) ban on news links in Canada.
I would be unsurprised if some product manager at Google finally decided to run the experiment and see if carrying Google News at all was worth the cost-benefit, or if dropping it would be more profitable (or even just easier).
posted by learning from frequent failure at 6:48 AM on February 23 [5 favorites]
Both Google and Facebook have had unhappy relationships with publishers for years, getting worse as publishers have lobbied governments to force the tech companies to pay license fees for links. When it becomes too expensive or they want to throw their weight around, the tech companies have been willing to block news sources. Witness Google’s eight-year ban on Google News in Spain, and Facebook’s continuing (?) ban on news links in Canada.
I would be unsurprised if some product manager at Google finally decided to run the experiment and see if carrying Google News at all was worth the cost-benefit, or if dropping it would be more profitable (or even just easier).
posted by learning from frequent failure at 6:48 AM on February 23 [5 favorites]
In related news, here's how Google is killing independent product review sites (indirectly, as it's more a case of big media publishers gaming Google to crowd out the sites that are bothering to test and evaluate stuff).
posted by rory at 6:50 AM on February 23 [3 favorites]
posted by rory at 6:50 AM on February 23 [3 favorites]
Yes, the shopping-type search results are a horror. The other day I was looking for something (not shopping, can't remember the topic), and somehow Google's algorithm misidentified me as shopping for whatever it was. There simply was no option to switch to another tab or whatever that would actually allow me to search for what I needed. I went to a topical website and found what I wanted, but it was disconcerting.
I keep running into this and it is incredibly frustrating. I was researching some different bicycles the other day, and even a search like "BikeBrand BikeModel wheelbase" would come up with a purely shopping result page, with no apparent way to get to a regular search page. It took a lot of work-arounds on search terms to not get the useless shopping pages. I hate it.
posted by Dip Flash at 7:12 AM on February 23 [4 favorites]
I keep running into this and it is incredibly frustrating. I was researching some different bicycles the other day, and even a search like "BikeBrand BikeModel wheelbase" would come up with a purely shopping result page, with no apparent way to get to a regular search page. It took a lot of work-arounds on search terms to not get the useless shopping pages. I hate it.
posted by Dip Flash at 7:12 AM on February 23 [4 favorites]
Haven’t seen news disappear but that would be a great shame, as it’s one of the ways I increase results quality. Of course the degree to which that works these days is pretty questionable as it’s subject to the same degradation the rest of Google is.
LLM bullshit is of course is just a huge quality minus across the board. I do not wish to be served up fake content whatsoever.
posted by Artw at 7:15 AM on February 23
LLM bullshit is of course is just a huge quality minus across the board. I do not wish to be served up fake content whatsoever.
posted by Artw at 7:15 AM on February 23
> Gmail could potentially be heading to the gallows in the near/mid future
i hadn't heard anything about this as i don't read any mainstream toxic bullshit platforms but apparently this was a viral hoax on hellsite 3. No, Google Isn’t Sunsetting Gmail [Gizmodo]
posted by glonous keming at 7:18 AM on February 23 [4 favorites]
i hadn't heard anything about this as i don't read any mainstream toxic bullshit platforms but apparently this was a viral hoax on hellsite 3. No, Google Isn’t Sunsetting Gmail [Gizmodo]
posted by glonous keming at 7:18 AM on February 23 [4 favorites]
It would certainly be the thing they could do to piss off people to the maximum in a single action, so never say never.
posted by Artw at 7:21 AM on February 23 [15 favorites]
posted by Artw at 7:21 AM on February 23 [15 favorites]
They datamine your email and you have to pay (now) for photo and file storage above a certain amount. Not sure it's in anyone's interest to sunset Gmail. Plain text is cheap to store.
posted by subdee at 7:27 AM on February 23 [1 favorite]
posted by subdee at 7:27 AM on February 23 [1 favorite]
Reached for comment over email Thursday, a Google spokesperson pointed Gizmodo to a one-line tweet from Gmail on X: “Gmail is here to stay.”OK, now I'm concerned.
posted by Not A Thing at 7:32 AM on February 23 [41 favorites]
Yesterday I couldn't get google to return more than a single page of results for any query. Flipped over to DDG which let me find what I was looking for. Figured it was some sort of A/B testing against my ad blocker or something but maybe it was more wide spread.
posted by Mitheral at 8:07 AM on February 23 [1 favorite]
posted by Mitheral at 8:07 AM on February 23 [1 favorite]
I do know that Google’s reverse image search has gotten progressively worse over the past year or so. A lot of the time TinEye actually works better (though even it has somehow gotten demonstrably worse over the past couple of weeks.)
posted by Thorzdad at 8:14 AM on February 23 [2 favorites]
posted by Thorzdad at 8:14 AM on February 23 [2 favorites]
GMail gets sold B2B (in competition with Microsoft), so I don't think it's going anywhere.
When I need to search news, I use DuckDuckGo. It gets the job done.
posted by humbug at 8:14 AM on February 23 [4 favorites]
When I need to search news, I use DuckDuckGo. It gets the job done.
posted by humbug at 8:14 AM on February 23 [4 favorites]
Find in page reveals that nobody has mentioned enshittification yet. This right here folks, is enshittification in action. And by this I mean "enshittification" not in the (increasingly popular, incorrect and meaning-robbing) sense of "something getting shittier over time," but the pure quill, the genuine article, the original sense(1).
(1) If anyone can point me at a more canonical original definition of "enshittification" I would like to be pointed at it. Thanks.
posted by Aardvark Cheeselog at 8:17 AM on February 23 [4 favorites]
(1) If anyone can point me at a more canonical original definition of "enshittification" I would like to be pointed at it. Thanks.
posted by Aardvark Cheeselog at 8:17 AM on February 23 [4 favorites]
Really was not expecting Google of all companies to make even the tiniest foray into the open source AI world despite the We Have No Moat memo.
I always found the “no moat” claims a bit surprising because while people are doing some impressive things with small models, it’s not really my impression that the benefits of scale have been disproven, and Google is probably top of the heap for capacity to do ML at scale? Of all the “pivots to AI” theirs should be one of the most plausible on paper - but they are such a mess on the product side these days.
posted by atoxyl at 8:20 AM on February 23 [1 favorite]
I always found the “no moat” claims a bit surprising because while people are doing some impressive things with small models, it’s not really my impression that the benefits of scale have been disproven, and Google is probably top of the heap for capacity to do ML at scale? Of all the “pivots to AI” theirs should be one of the most plausible on paper - but they are such a mess on the product side these days.
posted by atoxyl at 8:20 AM on February 23 [1 favorite]
atoxyl: fully agreed. They have Youtube, they have unparalleled hardware scale. And as AI Explained notes ~2:30 in his Sora breakdown almost all of the citations in OpenAI’s own technical report for Sora come from Google. There’s zero excuse for OpenAI eating their lunch so badly.
Morning reading was Gemma-7B reactions on the usual ML/open source AI subreddits: ranged from shrugs to fairly savage. Facebook benefits massively from their engagement with the community and it’s mutual, but the general vibe for this is that Google’s handing them scraps and saying “please polish our turd for free” because they’re losing to OpenAI. Zero enthusiasm.
Is this what class consciousness looks like in 2024?
At any rate, best comment was someone pointing out the irony that “Open”AI is now the lone fully closed-source major player in the field.
posted by Ryvar at 8:47 AM on February 23 [1 favorite]
Morning reading was Gemma-7B reactions on the usual ML/open source AI subreddits: ranged from shrugs to fairly savage. Facebook benefits massively from their engagement with the community and it’s mutual, but the general vibe for this is that Google’s handing them scraps and saying “please polish our turd for free” because they’re losing to OpenAI. Zero enthusiasm.
Is this what class consciousness looks like in 2024?
At any rate, best comment was someone pointing out the irony that “Open”AI is now the lone fully closed-source major player in the field.
posted by Ryvar at 8:47 AM on February 23 [1 favorite]
This month's Harper's has a story that points out a lot of the money flows associated with "AI" projects (very much including Google's) come from the US military and surveillance budgets.
The Pentagon’s Silicon Valley Problem
The author's POV is that, in a pattern in evidence since the Vietnam war, such systems have mostly failed to deliver promised results, but the government predictably keeps spending on them anyway, gradually subsuming tech businesses and tech culture into the old military industrial complex.
posted by Western Infidels at 8:50 AM on February 23 [7 favorites]
The Pentagon’s Silicon Valley Problem
The author's POV is that, in a pattern in evidence since the Vietnam war, such systems have mostly failed to deliver promised results, but the government predictably keeps spending on them anyway, gradually subsuming tech businesses and tech culture into the old military industrial complex.
posted by Western Infidels at 8:50 AM on February 23 [7 favorites]
Oh of course there’s a massive public subsidy in addition to the VC - I’d assumed it was all a loss leader effort to get it embedded everywhere before the rug pull, but if they can do that for free they are totally going to do it.
posted by Artw at 8:54 AM on February 23
posted by Artw at 8:54 AM on February 23
I find AI itself is enshittifying at a rapid rate. I have a recurring, tedious task in which I have to shift footnotes from other formats to Chicago format. In December, ChatGPT did this brilliantly. So I told all my friends I had found a tedious task ChatGPT was actually useful for.
Now, using the exact same prompts, it is generally incapable of putting footnotes in Chicago format. I thought it might be something about the new footnotes I received, so I went back and used my dataset from December. It couldn't do those either, even though it had performed flawlessly in December. It didn't even produce results that were internally consistent. I then tried other AIs with equally lousy results.
I'm hearing horror stories from friends at agencies where execs in their organizations say, "I used AI to produce this client strategy document in ten minutes. Look! It's brilliant!" And then the exec hands over 12 pages of word salad to present to the client.
"Pivot to video" killed heaps of news organizations. I think "use AI to speed things up" is going to take down companies generally, because the credulous use of AI will produce some pretty dire results.
posted by rednikki at 9:08 AM on February 23 [26 favorites]
Now, using the exact same prompts, it is generally incapable of putting footnotes in Chicago format. I thought it might be something about the new footnotes I received, so I went back and used my dataset from December. It couldn't do those either, even though it had performed flawlessly in December. It didn't even produce results that were internally consistent. I then tried other AIs with equally lousy results.
I'm hearing horror stories from friends at agencies where execs in their organizations say, "I used AI to produce this client strategy document in ten minutes. Look! It's brilliant!" And then the exec hands over 12 pages of word salad to present to the client.
"Pivot to video" killed heaps of news organizations. I think "use AI to speed things up" is going to take down companies generally, because the credulous use of AI will produce some pretty dire results.
posted by rednikki at 9:08 AM on February 23 [26 favorites]
I just noticed the other day that Google News no longer lets you do searches for results from a particular time range. You can say "the past day" or "the past year," but you can't do a search for, say, 2000-2005 anymore. Maybe there's still a way to edit the URL or use magic keywords to do it, I don't know.
It looks like they've also switched news results to some kind of infinite scroll, where you don't know how many results there are.
posted by smelendez at 9:13 AM on February 23 [3 favorites]
It looks like they've also switched news results to some kind of infinite scroll, where you don't know how many results there are.
posted by smelendez at 9:13 AM on February 23 [3 favorites]
Yes, Yahoo and no, I couldn't believe it myself that I'd fallen back on that ancient (by Internet standards) site...
I don't care if it's the newest or fanciest with bells, whistles, and bling, all I want is something up to date and fairly reliable.
Don't tinker with things, don't improve things, check to see the power is on and that the servers are running...
Except we monkeys are now conditioned to expect something new and shiny coming down the pike frequently to keep us amused.
I used Bing AI recently to good effect!
Again: I made sure to look up each detail in what Bing spat at me. And it was correct!
And why should you have to check every bit of info that comes down? What happens when you're in a hurry, or you trust Slimy Bing's Shitty AI to provide you with the info so you don't check, and there's one glaring error that screws the pooch?
I hate AI. I loath AI. AI is going to be the death of the internet, maybe even the death of a humane society. Call me Cassandra. I hope I'm wrong. AI could have been something moderately useful in a few years under the proper development. Not at this point, not where the AI juggernaut is rapidly heading. Misdirection, misinformation, slant, and outright lies, with just enough truth to suck people in.
I'm told this Wikipedia article sums things up well and has a balanced approach toward the subject. Eventually, there are lots of positive outcomes possible, but it's going to be a long time coming before AI could be truly useful and benefiting society. Reading carefully points out so many red flags. So much can, and looks like it is, starting to go wrong. We don't really even fully know how AI works let alone what it's capable of.
...artificial intelligence in the 21st century is influencing a societal and economic shift towards increased automation, data-driven decision-making, and the integration of AI systems into various areas of life, impacting job markets, healthcare, government, industry, and education. This raises questions about the ethical implications and risks of AI, prompting discussions about regulatory policies...
It is possible for bad actors to use this technology to create massive amounts of misinformation or propaganda.[156] AI pioneer Geoffrey Hinton expressed concern about AI enabling "authoritarian leaders to manipulate their electorates" on a large scale, among other risks.
Can you imagine what an intelligent sociopath with a Trump agenda could do with AI? Other countries seem to have the brakes on a bit better, but their government leanings are also not necessarily in the direction avoiding propaganda and working toward the betterment of society. Laws and directives are not actions to positively guide AI, they are substandard reactions to things that have gone wrong or an attempt to control bad actors.
AI is suposed to be working under...SUM values – developed by the Alan Turing Institute tests projects in four main areas:
RESPECT the dignity of individual people
CONNECT with other people sincerely, openly and inclusively
CARE for the wellbeing of everyone
PROTECT social values, justice and the public interest
We live in a worldwide corporatocracy influenced by multibillionaires. This is not how AI is going to be used. This genie should have never been let out of the bottle. We monkeys have never learned that just because we can do something, doesn't mean we should. It's got lots of noise and shine. So does a nuclear weapon.
Please read the article and come back and tell me exactly how wrong I am.
posted by BlueHorse at 9:17 AM on February 23 [6 favorites]
I don't care if it's the newest or fanciest with bells, whistles, and bling, all I want is something up to date and fairly reliable.
Don't tinker with things, don't improve things, check to see the power is on and that the servers are running...
Except we monkeys are now conditioned to expect something new and shiny coming down the pike frequently to keep us amused.
I used Bing AI recently to good effect!
Again: I made sure to look up each detail in what Bing spat at me. And it was correct!
And why should you have to check every bit of info that comes down? What happens when you're in a hurry, or you trust Slimy Bing's Shitty AI to provide you with the info so you don't check, and there's one glaring error that screws the pooch?
I hate AI. I loath AI. AI is going to be the death of the internet, maybe even the death of a humane society. Call me Cassandra. I hope I'm wrong. AI could have been something moderately useful in a few years under the proper development. Not at this point, not where the AI juggernaut is rapidly heading. Misdirection, misinformation, slant, and outright lies, with just enough truth to suck people in.
I'm told this Wikipedia article sums things up well and has a balanced approach toward the subject. Eventually, there are lots of positive outcomes possible, but it's going to be a long time coming before AI could be truly useful and benefiting society. Reading carefully points out so many red flags. So much can, and looks like it is, starting to go wrong. We don't really even fully know how AI works let alone what it's capable of.
It is possible for bad actors to use this technology to create massive amounts of misinformation or propaganda.[156] AI pioneer Geoffrey Hinton expressed concern about AI enabling "authoritarian leaders to manipulate their electorates" on a large scale, among other risks.
Can you imagine what an intelligent sociopath with a Trump agenda could do with AI? Other countries seem to have the brakes on a bit better, but their government leanings are also not necessarily in the direction avoiding propaganda and working toward the betterment of society. Laws and directives are not actions to positively guide AI, they are substandard reactions to things that have gone wrong or an attempt to control bad actors.
AI is suposed to be working under...SUM values – developed by the Alan Turing Institute tests projects in four main areas:
RESPECT the dignity of individual people
CONNECT with other people sincerely, openly and inclusively
CARE for the wellbeing of everyone
PROTECT social values, justice and the public interest
We live in a worldwide corporatocracy influenced by multibillionaires. This is not how AI is going to be used. This genie should have never been let out of the bottle. We monkeys have never learned that just because we can do something, doesn't mean we should. It's got lots of noise and shine. So does a nuclear weapon.
Please read the article and come back and tell me exactly how wrong I am.
posted by BlueHorse at 9:17 AM on February 23 [6 favorites]
Re. recent enshitification:
Reportedly ChatGPT is worse lately because the hidden “system” prompt has ballooned to 1700 “tokens” (~words) of caveats and guardrails that tend to distract and confuse from user intent. A workaround is to use the API playground, which circumvents some of this.
(Citations of evidence for this exist, but am presently on mobile.)
posted by Verg at 9:17 AM on February 23 [6 favorites]
Reportedly ChatGPT is worse lately because the hidden “system” prompt has ballooned to 1700 “tokens” (~words) of caveats and guardrails that tend to distract and confuse from user intent. A workaround is to use the API playground, which circumvents some of this.
(Citations of evidence for this exist, but am presently on mobile.)
posted by Verg at 9:17 AM on February 23 [6 favorites]
Western Infidel: I can vouch from personal experience that defense subcontractors have been recruiting game developers for the past few years to help assemble Unreal/Unity-based military drone piloting sims. Initially for humans utilizing VR controls but repurposing that for AI training would require almost zero additional work on the simulator side - in Unreal you’d swap out PlayerController for AIController and link your expert system to that, begin developing the AI side from there. Even in a cold-call-turned-impromptu-phone interview, AI being on the roadmap was hinted at and I assume with fairly high confidence this is already happening.
With the disclaimer that I fucking hate Good Will Hunting: my response wasn’t this good, but I’d like to think I left some ghouls with bruised egos. The initial salary range was twice what I’m making so somebody took the job.
rednikki: AI itself is enshittifying at a rapid rate
There’s been a ton of suspicion that OpenAI has been making drastic, overly aggressive changes to … on preview, Verg beat me to it. Then this happened a couple days ago, so.. yeah. See also: importance of open source AI because if Capital could feed us dogfood they’d feed us dogfood and this is no different. As long as the systems are centrally controlled any means of employing them for the benefit of workers will be strangled.
Related note: the risk of enshittification via model collapse is greatly overhyped, and if you’re waiting for that to strangle LLM development don’t hold your breath. Everyone working in the space is aware of the danger and very few people sufficiently immersed in the field to be making the fine-grained training data decisions are particularly stupid, though occasionally their bosses are.
posted by Ryvar at 9:50 AM on February 23 [4 favorites]
With the disclaimer that I fucking hate Good Will Hunting: my response wasn’t this good, but I’d like to think I left some ghouls with bruised egos. The initial salary range was twice what I’m making so somebody took the job.
rednikki: AI itself is enshittifying at a rapid rate
There’s been a ton of suspicion that OpenAI has been making drastic, overly aggressive changes to … on preview, Verg beat me to it. Then this happened a couple days ago, so.. yeah. See also: importance of open source AI because if Capital could feed us dogfood they’d feed us dogfood and this is no different. As long as the systems are centrally controlled any means of employing them for the benefit of workers will be strangled.
Related note: the risk of enshittification via model collapse is greatly overhyped, and if you’re waiting for that to strangle LLM development don’t hold your breath. Everyone working in the space is aware of the danger and very few people sufficiently immersed in the field to be making the fine-grained training data decisions are particularly stupid, though occasionally their bosses are.
posted by Ryvar at 9:50 AM on February 23 [4 favorites]
> the risk of enshittification via model collapse is greatly overhype
This. The only reason I think people are saying this is that they want it to be true.
The Internet will collapse, but the models will be fine. They'll just do different things than you expect.
posted by constraint at 10:01 AM on February 23 [3 favorites]
This. The only reason I think people are saying this is that they want it to be true.
The Internet will collapse, but the models will be fine. They'll just do different things than you expect.
posted by constraint at 10:01 AM on February 23 [3 favorites]
Google's news suggestions were always so terrible. I don't need to know about a new restaurant opening up in Duluth. I don't live in Duluth. I've never been there. Why is it showing me this? And then I'd get all those random radio stations reposting AP articles with their ads slapped on. I would diligently try to correct Google and say I don't want to see those things but it never made a difference. I've noticed over the past couple days I can't even tell it to avoid a source or a topic anymore. It's just god awful.
posted by downtohisturtles at 10:09 AM on February 23 [2 favorites]
posted by downtohisturtles at 10:09 AM on February 23 [2 favorites]
Then this happened a couple days ago
The article quotes someone calling it "nonsense, sometimes Shakespearean nonsense" (and, from the URL, it looks like the original headline reflected that, too) but lines like "Keep the choirs to the betide, brack, and blish, and eld the pruner with croid and tattle" are really much more Finnegans Wake-like -- kind of uncannily so, how it hits the rhythms you find all over FW.
posted by nobody at 10:23 AM on February 23 [6 favorites]
The article quotes someone calling it "nonsense, sometimes Shakespearean nonsense" (and, from the URL, it looks like the original headline reflected that, too) but lines like "Keep the choirs to the betide, brack, and blish, and eld the pruner with croid and tattle" are really much more Finnegans Wake-like -- kind of uncannily so, how it hits the rhythms you find all over FW.
posted by nobody at 10:23 AM on February 23 [6 favorites]
Reportedly ChatGPT is worse lately because the hidden “system” prompt has ballooned to 1700 “tokens” (~words) of caveats and guardrails that tend to distract and confuse from user intent.
Reminds me of when RoboCop became completely ineffective because OCP crammed hundreds of new directives into his program.
posted by Servo5678 at 10:23 AM on February 23 [6 favorites]
Reminds me of when RoboCop became completely ineffective because OCP crammed hundreds of new directives into his program.
posted by Servo5678 at 10:23 AM on February 23 [6 favorites]
We’re introducing a new offering called Gemini Business, which lets organizations use generative AI in Workspace at a lower price point than Gemini Enterprise, which replaces Duet AI for Workspace Enterprise.I think my favorite part of any google product announcement these days is that you could casually just drop the phrase "Bames Nond's having a stronk, call a Bondulance" into the middle of it and it would fit right in, like you'd barely notice it was there.
— Thomas Kurian (@ThomasOrTK) February 21, 2024
posted by mhoye at 10:26 AM on February 23 [19 favorites]
It really is amazing how much we are making calls to generative AI services a single point of failure in critical infrastructure, just because a bunch of money and hype men are telling us to, and without regard for if it is useful, appropriate or reliable. The consequences are likely to be awful.
posted by Artw at 10:40 AM on February 23 [13 favorites]
posted by Artw at 10:40 AM on February 23 [13 favorites]
Why on earth would Google shitcan gmail?
Can you think of a better / easier source of LLM training data?
posted by Sauce Trough at 10:52 AM on February 23 [4 favorites]
Can you think of a better / easier source of LLM training data?
posted by Sauce Trough at 10:52 AM on February 23 [4 favorites]
Artw: we don’t agree on this topic a lot so I really like it when we’re 100% agreed. Absolutely nothing based on LLMs should be ever be in that position, period.
posted by Ryvar at 10:54 AM on February 23 [2 favorites]
posted by Ryvar at 10:54 AM on February 23 [2 favorites]
I don’t think we are quite as often in disagreement as you think, you just tend to be more “this has cool and interesting applications if not used in a terrible way” and I lean more “Us living in the terrible society we do means THEY ARE TOTALLY GOING TO USE THIS IN THE TERRIBLE WAY”, is all.
posted by Artw at 11:15 AM on February 23 [14 favorites]
posted by Artw at 11:15 AM on February 23 [14 favorites]
Why can't we put "NO" on the "Generate AI bullshit on this search" - only offers "generate". Nah, give me an explicit "fuck you and your bullshit" - but you cowards would rather force shit on end users and pretend its our "choice".
Also fuck them for hiding news. If there's at least one half way decent thing left it's getting *some* news outta the damn thing. Image search results suck, regular results have been shittified for years, and only getting worse.
posted by symbioid at 11:18 AM on February 23
Also fuck them for hiding news. If there's at least one half way decent thing left it's getting *some* news outta the damn thing. Image search results suck, regular results have been shittified for years, and only getting worse.
posted by symbioid at 11:18 AM on February 23
Why on earth would Google shitcan gmail?
Can you think of a better / easier source of LLM training data?
Why have people feed your LLM training corpus for free when you can get a few bucks a month AND have them all train the corpus, because what are they gonna do, buy their own domain and run their own mail server?
posted by tclark at 11:31 AM on February 23 [6 favorites]
Can you think of a better / easier source of LLM training data?
Why have people feed your LLM training corpus for free when you can get a few bucks a month AND have them all train the corpus, because what are they gonna do, buy their own domain and run their own mail server?
posted by tclark at 11:31 AM on February 23 [6 favorites]
Thanks I hate it.
posted by lock robster at 11:58 AM on February 23
posted by lock robster at 11:58 AM on February 23
Arg. I've used Google News daily for years, both the front page and for search.
Is anything comparable in the front page side?
posted by doctornemo at 12:01 PM on February 23 [1 favorite]
Is anything comparable in the front page side?
posted by doctornemo at 12:01 PM on February 23 [1 favorite]
Can you think of a better / easier source of LLM training data?
99.9% of the email I get is spam. I suspect I'm a fairly typical user of their gmail product. If Google is training their models to overfit on garbage, then investors should probably ask some very hard questions about the quality of Google's AI products.
posted by They sucked his brains out! at 12:22 PM on February 23 [2 favorites]
99.9% of the email I get is spam. I suspect I'm a fairly typical user of their gmail product. If Google is training their models to overfit on garbage, then investors should probably ask some very hard questions about the quality of Google's AI products.
posted by They sucked his brains out! at 12:22 PM on February 23 [2 favorites]
When they changed Google news a while back, I switched to newsnow.com.
posted by The Half Language Plant at 1:17 PM on February 23 [2 favorites]
posted by The Half Language Plant at 1:17 PM on February 23 [2 favorites]
99.9% of the email I get is spam.
That's OK, that's 99.9% of what LLMs are used for.
posted by mhoye at 1:29 PM on February 23 [8 favorites]
That's OK, that's 99.9% of what LLMs are used for.
posted by mhoye at 1:29 PM on February 23 [8 favorites]
Wait, is there even news on the internet anymore?
Kidding, but only kinda. Considering that most of the decent news sources are either dying out (RIP DCist, one of the last publications covering local DC news, murdered today) or are behind paywalls, the only open so-called "news" sources left are basically either AI drool or right-wing nut-job conspiracy sites. Maybe Google killing "news" searches is good news.
posted by General Malaise at 2:00 PM on February 23 [3 favorites]
Kidding, but only kinda. Considering that most of the decent news sources are either dying out (RIP DCist, one of the last publications covering local DC news, murdered today) or are behind paywalls, the only open so-called "news" sources left are basically either AI drool or right-wing nut-job conspiracy sites. Maybe Google killing "news" searches is good news.
posted by General Malaise at 2:00 PM on February 23 [3 favorites]
fair cop
posted by They sucked his brains out! at 2:00 PM on February 23
posted by They sucked his brains out! at 2:00 PM on February 23
Just now I searched for a store *by name* to try to find its official website. Had to settle for clicking its wikipedia page and scrolling down to the "official website" link in the wiki article. Neither Google or DuckDuckGo would show it in the first few pages of results.
posted by ctmf at 2:30 PM on February 23 [3 favorites]
posted by ctmf at 2:30 PM on February 23 [3 favorites]
It's entering "Do you even use your own product? And you're happy with this?" territory. I guess if it makes money, then the answer could be yes, but pretending it's about being a useful service and not just getting people to serve ads to themselves is increasingly not credible.
posted by ctmf at 4:19 PM on February 23 [2 favorites]
posted by ctmf at 4:19 PM on February 23 [2 favorites]
It feels like there's been a very fast decline in Google quality. Or is that just me? For years I'd grumble about this or that little problem with Google search, but it was still a very useful product. But over the past couple of months... I dunno, I'm not sure if it just crossed my personal line of tolerance or if most people are having similar feelings. Not even the past year, just the past couple of months.
posted by clawsoon at 4:40 PM on February 23 [8 favorites]
posted by clawsoon at 4:40 PM on February 23 [8 favorites]
It's funny to read the first comment on the oldest story in "Related Posts":
Wow, I have this odd romance for Google. They are to the Internet as Habitat For Humanity to communities.posted by clawsoon at 4:42 PM on February 23 [3 favorites]
I hope it's fucking embarrassing to be a Google engineer.
posted by ctmf at 4:42 PM on February 23 [1 favorite]
posted by ctmf at 4:42 PM on February 23 [1 favorite]
Thank you, Half Language Plant. Trying it out now.
posted by doctornemo at 4:59 PM on February 23
posted by doctornemo at 4:59 PM on February 23
As I said up top: Yahoo News.
I need Canadian news and most of the recommended news aggregators focus on US news but https://ca.news.yahoo.com/ gives me at least some Canadian content. Yes, it's annoying that every fourth or fifth story is an ad for some product I'd never purchase or some clickbaity celebrity "news" but it's about the best replacement I could find for Google News. Now, as you're in the US, I wouldn't suggest the Canadian site, but the American one should be very similar.
As suggested in this Ask (which I posted looking for advice with this exact problem), MSN might also be a viable option.
posted by sardonyx at 5:37 PM on February 23
I need Canadian news and most of the recommended news aggregators focus on US news but https://ca.news.yahoo.com/ gives me at least some Canadian content. Yes, it's annoying that every fourth or fifth story is an ad for some product I'd never purchase or some clickbaity celebrity "news" but it's about the best replacement I could find for Google News. Now, as you're in the US, I wouldn't suggest the Canadian site, but the American one should be very similar.
As suggested in this Ask (which I posted looking for advice with this exact problem), MSN might also be a viable option.
posted by sardonyx at 5:37 PM on February 23
It feels like there's been a very fast decline in Google quality. Or is that just me? For years I'd grumble about this or that little problem with Google search, but it was still a very useful product. But over the past couple of months... I dunno, I'm not sure if it just crossed my personal line of tolerance or if most people are having similar feelings. Not even the past year, just the past couple of months.
They are laying off tens of thousands of employees. Might be related.
posted by ryoshu at 8:00 PM on February 23 [3 favorites]
They are laying off tens of thousands of employees. Might be related.
posted by ryoshu at 8:00 PM on February 23 [3 favorites]
Thank you, sardonyx. I'm amazed Yahoo hasn't killed it yet.
posted by doctornemo at 8:40 PM on February 23 [1 favorite]
posted by doctornemo at 8:40 PM on February 23 [1 favorite]
You know how the 20th century model of dictatorship is that the dictator comes in & then shuts down all the independent newspapers and press outlets? It seems like the 21st century version is that Silicon Valley and the venture capitalists are shutting down the press outlets first because they're betting an authoritarian regime will be voted in this November.
posted by jonp72 at 8:58 PM on February 23 [5 favorites]
posted by jonp72 at 8:58 PM on February 23 [5 favorites]
Why on earth would Google shitcan gmail?
They're not going to shitcan gmail entirely, but I fully expect that it's not going to be free for much longer. They have lots of data around what drives a user to purchase their own domain and turn Gsuite on for 2-3 users. A significant portion of paid-Gmail/gsuite usage is not corporate, already.
The users that Google wants to retain (and show advertising to, and train their models with) are the kind of users who don't mind shelling out $75-80/year for managed email on a custom domain. When @gmail.com addresses start to cost $30/year (maybe with the first 6 mo free), a significant portion of the userbase will immediately pay up -- it's a small cost to pay when compared to the pain of moving your primary email (or tolerating the 2000-era quality of your ISP's "free" offering).
posted by toxic at 10:37 AM on February 25
They're not going to shitcan gmail entirely, but I fully expect that it's not going to be free for much longer. They have lots of data around what drives a user to purchase their own domain and turn Gsuite on for 2-3 users. A significant portion of paid-Gmail/gsuite usage is not corporate, already.
The users that Google wants to retain (and show advertising to, and train their models with) are the kind of users who don't mind shelling out $75-80/year for managed email on a custom domain. When @gmail.com addresses start to cost $30/year (maybe with the first 6 mo free), a significant portion of the userbase will immediately pay up -- it's a small cost to pay when compared to the pain of moving your primary email (or tolerating the 2000-era quality of your ISP's "free" offering).
posted by toxic at 10:37 AM on February 25
When @gmail.com addresses start to cost $30/year (maybe with the first 6 mo free), a significant portion of the userbase will immediately pay up -- it's a small cost to pay when compared to the pain of moving your primary email...
This.
Gmail is so deeply entwined in people’s world and their identities, that to suddenly not have the Gmail address (or addresses) that you’ve used for a couple of decades would be a huuuuuuuuuuge problem. That, and there really isn’t anywhere else to go, realistically. And, no, nearlyfreespeech isn’t a viable option for most.
posted by Thorzdad at 12:48 PM on February 25
This.
Gmail is so deeply entwined in people’s world and their identities, that to suddenly not have the Gmail address (or addresses) that you’ve used for a couple of decades would be a huuuuuuuuuuge problem. That, and there really isn’t anywhere else to go, realistically. And, no, nearlyfreespeech isn’t a viable option for most.
posted by Thorzdad at 12:48 PM on February 25
Am I going to end up feeling smug for having stuck with Yahoo email for all these years?
posted by clawsoon at 1:31 PM on February 25 [1 favorite]
posted by clawsoon at 1:31 PM on February 25 [1 favorite]
That, and there really isn’t anywhere else to go, realistically.I switched from Gmail to Fastmail and I’ve been happy with it, FWIW.
posted by april of time at 9:18 PM on February 25 [1 favorite]
Fastmail is a better service, and it’s the work of minutes to switch and migrate your data, but it’s not free and Google knows most Gmail users don’t use custom domains and would need to get everyone they know to update their addresses. That’s what keeps Google executives sleeping at night knowing that they will be getting that sweet ad data for decades to come.
posted by adamsc at 4:31 AM on February 26 [1 favorite]
posted by adamsc at 4:31 AM on February 26 [1 favorite]
« Older “The most neglected of American fruits” | Moon landings, a wooden satellite, Tolkien on Mars... Newer »
This thread has been archived and is closed to new comments
posted by chavenet at 4:33 AM on February 23