(Thread IKs:
Platystemon)
|
militarized A.I. is a complete pipedream at this point for this planet's bourgeoisie #Baankrupt.
|
#
?
Apr 3, 2024 20:41
|
|
|
|
| # ? May 20, 2024 19:08 |
|
|
|
|
#
?
Apr 3, 2024 20:41
|
|
|
Ham Equity posted:"This AI thing is a huge time saver, just make sure you verify everything it tells you with other sources..." "have a great weekend!" -- liberals, before ending the call and turning off the lights
|
|
#
?
Apr 3, 2024 21:15
|
|
|
|
|
#
?
Apr 3, 2024 21:47
|
|
|
 from 1923 spechtie has issued a correction as of 22:31 on Apr 3, 2024 |
|
#
?
Apr 3, 2024 22:29
|
|
|
spechtie posted:
But did they consider the vast wealth of anime we would have constant access to? Future's not so bad now, huh?
|
|
#
?
Apr 3, 2024 22:34
|
|
|
spechtie posted:
little did they know the only people calling you would be robots who are very concerned about your car warranty, because everyone just texts instead
|
|
#
?
Apr 3, 2024 22:36
|
|
|
blastron posted:Reading the article, there really was an algorithm processing all of that data and identifying people’s purchases. The Indian workers were there to cover for when the algorithm wasn’t able to fully handle everything. As designed, intervention was only supposed to be necessary in 50/1000 of cases. how the gently caress does an algorithm mess up adding things lmao hahaha what
|
|
#
?
Apr 4, 2024 00:34
|
|
|
|
|
#
?
Apr 4, 2024 00:48
|
|
|
https://x.com/Reuters/status/1775655880127574321?s=20 wow seems smart
|
|
#
?
Apr 4, 2024 06:59
|
|
|
Nichael posted:https://x.com/Reuters/status/1775655880127574321?s=20 they will make it work by completely nuking the free one and rendering it even more useless
|
|
#
?
Apr 4, 2024 07:30
|
|
|
Take the plunge! Okay! posted:they will make it work by completely nuking the free one and rendering it even more useless Bing is back baby!
|
|
#
?
Apr 4, 2024 08:28
|
|
|
Take the plunge! Okay! posted:they will make it work by completely nuking the free one and rendering it even more useless And I thought that would stop the garbage AI results appearing at search.
|
|
#
?
Apr 4, 2024 08:43
|
|
|
Nichael posted:https://x.com/Reuters/status/1775655880127574321?s=20 kagi search is becoming more viable by the day
|
|
#
?
Apr 4, 2024 13:58
|
|
|
Nichael posted:https://x.com/Reuters/status/1775655880127574321?s=20 I feel like this is a soft way to drop it completely because I've been in the experimental trial of this. It sucks poo poo, and they can't sell ads on it.
|
|
#
?
Apr 4, 2024 14:26
|
|
|
No the plan is to make regular search so unusable and inundated with ai poo poo that you have to use their paid search to find real poo poo.
|
|
#
?
Apr 4, 2024 14:29
|
|
|
If a piece of information isn't on Wikipedia or wikiHow it effectively doesn't exist
|
|
#
?
Apr 4, 2024 14:36
|
|
|
Heck Yes! Loam! posted:No the plan is to make regular search so unusable and inundated with ai poo poo that you have to use their paid search to find real poo poo. nah, paid chatbots that may or may not be lying to you
|
|
#
?
Apr 4, 2024 14:38
|
|
|
that's like an incel starting to charge for sex
|
|
#
?
Apr 4, 2024 14:41
|
|
|
I still don't get how a lot of the companies putting unbelievable amounts of money into AI expect to make money off it. Adobe and Microsoft I can at least see some way as they at least look to be adding it to preexisting paid for services they offer so there's at least like a path for it -although MS 100 billion investment in a dataceter for ai seems insane- but others it really seems to be pour cash into it now and figure out the fiances of it later. I guess that is pretty common thing in tech investing, but it always seems a bit weird. Also has anyone ever actually done a study into how much money is made vs lost in all the large investments into new tech start ups? dr_rat has issued a correction as of 14:59 on Apr 4, 2024 |
|
#
?
Apr 4, 2024 14:57
|
|
|
I think it's also "if we aren't announcing dumping vast sums of money into the latest trend our shareholders will get upset"
|
|
#
?
Apr 4, 2024 15:00
|
|
|
dr_rat posted:I still don't get how a lot of the companies putting unbelievable amounts of money into AI expect to make money off it. Adobe and Microsoft I can at least see some way as they at least look to be adding it to preexisting paid for services they offer so there's at least like a path for it -although MS 100 billion investment in a dataceter for ai seems insane- but others it really seems to be pour cash into it now and figure out the fiances of it later. Every federal agency is being required to hire an CAIO (Chief AI Officer) so the usual path to profitability of “overcharge the government for things that don’t work” is wide open for AI tech. Plus OpenAI, Google, Amazon, etc are trying to position themselves on the selling shoves side of the gold rush. Let a thousand startups crash and burn and they’ll be the ones hoovering up the cash.
|
|
#
?
Apr 4, 2024 15:11
|
|
|
ArmedZombie posted:kagi search is becoming more viable by the day
|
|
#
?
Apr 4, 2024 15:14
|
|
|
Heck Yes! Loam! posted:No the plan is to make regular search so unusable and inundated with ai poo poo that you have to use their paid search to find real poo poo. They made $240 BILLION on advertising on their search pages last year. They are not going to intentionally kill that for an AI play. Putting the AI search behind a paywall is them drowning it in the bathtub. It'll get no revenue and they'll decommission it in 2 years. They do not intend for it to be poo poo; they are just incapable of keeping it from being poo poo and also making money off it. They're shoving ads in wherever they possibly can, which makes the product worse. They've also made a ton of core search algorithm changes aimed at punishing people trying to "trick" Google into giving them good rankings with SEO tactics. This is the part they're failing miserably at. They put out a bunch of guidelines that are like, "Don't write content for Google; write it for users, and we will reward content that helps users," but it's really obvious from all the rankings that they have not figured out how to give good rankings to good content and punish bad content.
|
|
#
?
Apr 4, 2024 15:36
|
|
|
gee who would have thought... Washington's Lottery AI site turned user photo into porn mynorthwest.com - Wed, 03 Apr 2024 posted:A new Washington’s Lottery AI mobile site turned a user’s photo into softcore pornography, forcing them to take the website down “out of an abundance of caution.”
|
|
#
?
Apr 4, 2024 16:04
|
|
|
can't fire the computer oh drat the computer also says my rent has to go up, oh well i can't have surgery? well, the computer speaketh
|
|
#
?
Apr 4, 2024 16:09
|
|
|
PostNouveau posted:They've also made a ton of core search algorithm changes aimed at punishing people trying to "trick" Google into giving them good rankings with SEO tactics. This is the part they're failing miserably at. They put out a bunch of guidelines that are like, "Don't write content for Google; write it for users, and we will reward content that helps users," but it's really obvious from all the rankings that they have not figured out how to give good rankings to good content and punish bad content. That last part is something I think about all the time. Google has an incentive to make their search work - it's the foundation of their economic legitimacy - but it's a cat and mouse game between them and SEO hucksters, and the SEO hucksters have been winning, and winning hard, the last couple years. LLMs have if anything advantaged the SEO hucksters more than any other group, since "grammatically coherent bullshit" is not useful for a lot of things but it is useful for getting highly ranked in algorithmic search engines. And yeah it doesn't make much sense for google to pivot from something with low costs and high value to something with high costs and low value. Companies sometimes do that but it's usually smarter to assume they like making a profit.
|
|
#
?
Apr 4, 2024 22:53
|
|
|
Tulip posted:That last part is something I think about all the time. Google has an incentive to make their search work - it's the foundation of their economic legitimacy - but it's a cat and mouse game between them and SEO hucksters, and the SEO hucksters have been winning, and winning hard, the last couple years. LLMs have if anything advantaged the SEO hucksters more than any other group, since "grammatically coherent bullshit" is not useful for a lot of things but it is useful for getting highly ranked in algorithmic search engines. Well, the problem of SEO hucksters is like 4 or 5 years ago you could rank for just having more words about a subject. But the issue now seems to be that in an effort to root out the SEO hucksters, Google's algorithm is just garbage. It's not that you get SEO huckster results, you get loving incoherent results now. They pulled some important components out of their hugely complex machine and now it barely functions. All the school zone lights around me are off when they should be on, so I googled when spring break is for the local schools and the first result was a loving TripAdvisor article from 2018.
|
|
#
?
Apr 5, 2024 00:12
|
|
|
lol that all the biggies pivoted so seamlessly into this stuff from crypto's bloated corpse
|
|
#
?
Apr 5, 2024 03:11
|
|
|
there's always a new grift
|
|
#
?
Apr 5, 2024 03:12
|
|
|
All our male developers tested the revenge porn generator repurposed into a website widget and none of them got drawn with tits out, seems good enough
|
|
#
?
Apr 5, 2024 06:38
|
|
|
JAY ZERO SUM GAME posted:can't fire the computer well number was already running everything and sentencing people to death, they're just giving number the ability to talk
|
|
#
?
Apr 5, 2024 06:45
|
|
|
dr_rat posted:I still don't get how a lot of the companies putting unbelievable amounts of money into AI expect to make money off it. Adobe and Microsoft I can at least see some way as they at least look to be adding it to preexisting paid for services they offer so there's at least like a path for it -although MS 100 billion investment in a dataceter for ai seems insane- but others it really seems to be pour cash into it now and figure out the fiances of it later. the trajectory of “AI” in the last few years genuinely has been amazing so they all assume they’re going to make a ton on their investments and don’t consider that we’ve already hit a plateau 
|
|
#
?
Apr 5, 2024 14:53
|
|
|
eSports Chaebol posted:the trajectory of “AI” in the last few years genuinely has been amazing so they all assume they’re going to make a ton on their investments and don’t consider that we’ve already hit a plateau Eh, I don't think we've hit that plateau just yet, there's still lots of regular improvements happening to the current big AI's, but also I don't think we're anywhere near the level of what a lot of what's being promised AI can do, or what corporations are wanting it to do. AI Chat still at heart has no idea what it's output means and there's only so much clever programing and hacks you can do to hide that. Adobe seems smart on this, as they seem to just be offering it as a powerful image tool, but one that the end user is still expected to look over and correct and modify to make the output fit for end use. Well that's exactly the level of AI were at now, so they don't even have to over promise anything.
|
|
#
?
Apr 5, 2024 15:38
|
|
|
my personal hot take is that an LLM probably really could be AI with one small tweak: allow it to dynamically recalculate the weights in its model based on user input. hey it would probably only take a few terawatts of continuous power to do this today!
|
|
#
?
Apr 5, 2024 15:41
|
|
|
eSports Chaebol posted:my personal hot take is that an LLM probably really could be AI with one small tweak: allow it to dynamically recalculate the weights in its model based on user input. hey it would probably only take a few terawatts of continuous power to do this today! Look pretty sure fission reactors will be perfected any day now, so might as well just class that extreme power use as a solved problem. Lets start shipping this tomorrow! Edit: Wait we need to be realistic about this. Tomorrows a saturday, ship it on monday.
|
|
#
?
Apr 5, 2024 15:52
|
|
|
mawarannahr posted:I pay for it and it sucks rear end that's because you're searching the internet, which is bad now
|
|
#
?
Apr 5, 2024 16:43
|
|
|
eSports Chaebol posted:my personal hot take is that an LLM probably really could be AI with one small tweak: allow it to dynamically recalculate the weights in its model based on user input. hey it would probably only take a few terawatts of continuous power to do this today! I read all of the wikipedia articles about the AIs in the Halo video games, and I think youre on to something
|
|
#
?
Apr 5, 2024 18:45
|
|
|
Biplane posted:I read all of the wikipedia articles about the AIs in the Halo video games, and I think youre on to something it is canonical that john halo's suit sucks him off
|
|
#
?
Apr 5, 2024 19:06
|
|
|
|
| # ? May 20, 2024 19:08 |
|
|
eSports Chaebol posted:my personal hot take is that an LLM probably really could be AI with one small tweak: allow it to dynamically recalculate the weights in its model based on user input. hey it would probably only take a few terawatts of continuous power to do this today! "How do you dynamically recalculate weights in a useful way" is basically just saying make it an AI. LLMs are not, and will not, be true GAI.
|
|
#
?
Apr 5, 2024 19:09
|
|
