Not surprised. The CEO announced they plan to restrict access to logged-in users because AI Overviews on Google are killing traffic. Clearly they hope that'll lead to increased activity, but I suspect it'll have a modest initial positive impact and then lead to a slump in overall visits in the long- to very-long term. Not great. Their shitty $60M/year deal with Google for access to posts for training AI is also up, and they're negotiating renewal. They never should've sold access for chump change like that - that's Google's budget for disposable coffee cups. Google give Mozilla $400+ million each year just to retain Google as the default search engine in FF; Reddit sold access to some of the most lucrative training data available on the internet for pennies.
People aren't clicking on the site because the AI overview is summarising topics for users... the solution should be to make it impossible to load the site!
Anonymous A (OP) triple-posted this , 2 minutes later, 18 minutes after the original post[^][v]#1,468,884
Ngl, I kinda hope at least one tech company dies from this whole AI thing. They want to replace workers, but it would be a form of karma if at least some of them end up killing themselves in the process.
Anonymous C replied with this , 2 minutes later, 20 minutes after the original post[^][v]#1,468,886
@1,468,878 (A)
The whole internet is fucked. Google is slowly strangling it (and itself!) by killing traditional search and sites like Reddit are too busy heaping up their own funeral pyres.
Anonymous C double-posted this , 1 minute later, 22 minutes after the original post[^][v]#1,468,887
@1,468,879 (Oatmeal Fucker !BYUc1TwJMU) > People aren't clicking on the site because the AI overview is summarising topics for users... the solution should be to make it impossible to load the site!
Top minds at Reddit, clearly. Top. Minds.
Anonymous A (OP) replied with this , 1 minute later, 24 minutes after the original post[^][v]#1,468,889
@1,468,886 (C)
Early on when google added AI responses, people were saying, "Wait but if AI is trained on the internet but people stop clicking on websites because they’re using AI instead, why will people write websites, and if the websites are AI generated content how will the AI get better?"
It’s basically the internet committing suicide at this point.
Anonymous A (OP) double-posted this , 1 minute later, 25 minutes after the original post[^][v]#1,468,890
And of course, with OpenAI and Anthropic, AI won’t go away, but you can train your own LLM on the output of frontier LLMs the same way frontier LLMs were trained on the internet and create high quality AI models for a fraction of the price. So the end result is I think the whole tech industry is just putting a bullet in its own foot.
Anonymous C replied with this , 32 seconds later, 27 minutes after the original post[^][v]#1,468,893
@1,468,889 (A)
Exactly. It's GIGO. Garbage In, Garbage Out. Eventually LLMs will face model collapse, because we already know LLMs trained on enough synthetic data become retarded.
Anonymous A (OP) replied with this , 2 minutes later, 30 minutes after the original post[^][v]#1,468,894
@previous (C)
To be fair, there will be a moment when the tech peaks, and those models won’t go away… if we had access to them, but they’re proprietary. My hope is that since these models were trained on the stolen works of all of us, that we would regain collective ownership of AI models and everybody could just download a frontier AI model for free, and we have some law that you can’t sell AI. Probably won’t happen. But that would be the least bad path forward. The idea that one company is going to steal everything everyone ever created and sell it back to us for a subscription for the sake of causing mass unemployment while yapping about creating a conscious machine god and asking for trillions of dollars is kinda infuriating though.
Anonymous C replied with this , 56 seconds later, 31 minutes after the original post[^][v]#1,468,895
Stack Overflow is a great illustration of how fucked things are. It was one of the best repositories of programming questions and answers for LLMs, but nobody is using it anymore - everyone heads to AI. When models begin to degrade is anyone's guess, but it seems inevitable. They'll either collapse or be forced to stagnate, frozen in time for lack of human-authored training data.
Anonymous A (OP) replied with this , 1 minute later, 33 minutes after the original post[^][v]#1,468,896
@previous (C)
I noticed that. There used to be a lot of programming done to make programming better. But it seems like all that activity stopped and they’re just defaulting to building react sites recycling the same react components (or maybe the model is generating it from scratch watch time, but they look like the same components either way).
Anonymous A (OP) double-posted this , 1 minute later, 34 minutes after the original post[^][v]#1,468,897
Ngl, I think the best bet for the future might be physical forms of engineering. But that’s annoying because they have a higher barrier to entry. Not a fan of Mark Zuckerberg, but he wasn’t going to start the next Boeing out of a dorm room lol. Software was kinda cool in that regard.
Anonymous A (OP) triple-posted this , 2 minutes later, 37 minutes after the original post[^][v]#1,468,898
Idk, the AI thing already crashed in South Korea and in China. The US took it way further than anybody else though, we literally have 75% of the world’s processing power lol
> The whole internet is fucked. Google is slowly strangling it (and itself!) by killing traditional search and sites like Reddit are too busy heaping up their own funeral pyres.
It’s like locust behavior, over feeding until the seeds in the field are eaten too.
> Ngl, I think the best bet for the future might be physical forms of engineering. But that’s annoying because they have a higher barrier to entry. Not a fan of Mark Zuckerberg, but he wasn’t going to start the next Boeing out of a dorm room lol. Software was kinda cool in that regard.
I ran a simulation on a surviving Roman Empire. Currently, we’re not in a digital age but the level of engineering is ridiculous.
Live&Lemon joined in and replied with this , 6 minutes later, 1 hour after the original post[^][v]#1,468,938
@previous (E)
If I was teleported back to Ancient Rome and somehow I could speak Latin, I’d ask around, figure out how I could work with the guys that were making the weaponry for the Roman military and invent the steam engine.
Anonymous G joined in and replied with this , 2 hours later, 4 hours after the original post[^][v]#1,468,951
Apparently people stopped buying computers because AI companies made the components more expensive by buying everything so now Intel’s stock is tanking because they make computer chips.
Anonymous I joined in and replied with this , 20 minutes later, 4 hours after the original post[^][v]#1,468,958
jews: create AI to monetise things they have no ownership over
video game industry: you do not own the things you buy
Internet: You know what, fuck you both! *uses AI to recompile all the video games into a format thats easier to pirate*
jews and AAA video game developers: REEEEEEEEEEEEEEEEEEEEEEEEEE!!!1!
me:LOL
Live&Lemon joined in and replied with this , 12 minutes later, 19 hours after the original post[^][v]#1,469,115
@previous (K)
Nvidia is down 4% today… this is my guess but I could be wrong:
Nvidia is essentially selling shovels during a gold rush. Nvidia’s stock price might go down based on fears about an AI bubble, but that is probably an irrational reaction since Nvidia will still be enormously profitable until the AI companies (gold miners in this analogy) finally go bust. Then after they go bust, if and when they do, Nvidia will still have a business, just a smaller business, but they survived for years and years selling GPUs to gamers, so I think Nvidia might crash eventually, but probably won’t die.
Anonymous K replied with this , 3 minutes later, 19 hours after the original post[^][v]#1,469,118
@previous (Live&Lemon)
Good analysis, sensible and I agree. If they're wise they'll have planned for this happening and prepared for their company's future. Even without AI like you said we're always gonna need GPU's.
Live&Lemon replied with this , 2 minutes later, 19 hours after the original post[^][v]#1,469,119
But if Nvidia does crash, software companies (especially pure AI companies - Anthropic and OpenAI) should crash first and crash much harder. I think the problem with Anthropic is, when they released their S-1, I didn’t read the whole thing, but I sort of skimmed through it, and I looked to see what people were saying about it. The leaked information that came out weeks before the S-1 about the confidential S-1 pretty much had all the information that’s actually in the public S-1. So I don’t really trust that Anthropic is painting an accurate portrayal of their business, there could be numbers that they intentionally omitted because they don’t look good. Or maybe the numbers look great, I don’t really know. But investors will be investing on what they think is the case not what’s actually the case. So, that IPO could flop, or, alternatively, and maybe more likely, it doesn’t flop, and this whole thing gets even bigger. Which is probably the most concerning outcome, because these AI CEOs have a mindset where they’re obsessed with this ideological idea of a singularity or AGI where we have "super intelligence" and it solves all of humanity’s problems. If Anthropic’s IPO actually gets a $2 trillion valuation, Amodei might be a smart guy when it comes to computer science or neuroscience, but I would be willing to bet if he sees a sign that rapid success might be possible, he’s probably just going to continue the same cycle he’s been on and assume it’s going to continue forever and invest more money in more data centers and he’s going to eventually manifest an unsustainable financial situation, it will just be much bigger.
Live&Lemon double-posted this , 4 minutes later, 19 hours after the original post[^][v]#1,469,123
Like no matter how smart these billionaires are, they’re still human, and if you honest to God believe you’re are the center of the biggest revolution in technology in the history of humanity, and your company gets a $2 trillion valuation, no matter how restrained he might want to be, based on how billionaires tend to act, I really doubt he wouldn’t double down on it. I don’t think one day he’s just going to stop trying to push the frontier, but if Anthropic never becomes profitable, that would be the reason why. I could be wrong, nobody can predict the future.
(Basically the whole thing I think is super high risk high reward, but IMO the risk of getting it wrong and the losses associated with it make it not worth it even if later I find out a bunch of people get rich off of it)
Lawful Hungry !mjoQJhzjwM joined in and replied with this , 1 hour later, 20 hours after the original post[^][v]#1,469,133
@1,468,895 (C)
I mean it makes sense. Why ask noob questions on a site that has lots of angry programmers who are tired of answering the same question when you can fire up ChatGPT or Claude have it answer all sorts of questions and with provide you with some suggestions or ideas all in real time effectively eliminating necroposting altogether.
Anonymous C replied with this , 53 minutes later, 21 hours after the original post[^][v]#1,469,142
@previous (Lawful Hungry !mjoQJhzjwM)
It does make sense. It's the classic tragedy of the commons: when everyone acts in their own self-interest, it eventually works against the interests of everyone. The individual user reaps the short-term benefit but users collectively bear the long-term cost.
When I have a programming or tech question, I head to Claude, not Stack Overflow. I don't blame people, but I'm still unhappy with the situation, despite contributing to the problem myself. It's a strange feeling. Very occasionally if an LLM gives me a quick, valid solution to a difficult or obscure problem, and I find a thread on Reddit where nobody's been able to solve it, I'll chime in with the solution (assuming the post hasn't been archived or locked), but I don't go any further than that.
Anonymous C double-posted this , 3 minutes later, 21 hours after the original post[^][v]#1,469,143
@1,468,897 (A)
Definitely. Hardware and hands-on engineering, and physical human-computer interfaces, are going to be where a lot of people go. That's my prediction. The software space post-LLM isn't big enough to accommodate everyone with a passion for technology and building things. It's barely big enough to sustain the existing workforce as is.