Most likely the next step change in AI improvement will come from the next generation learning from today's LLMs and diffusion models. But not being trained on their output. More likely they will be evaluated by current AI.
AI advancements have a predictable pattern. First we mimic how humans do something, then we use those AIs to evaluate the next generation which learns how to do it without mimicking humans. LLMs have been trained on human output and work by trying to determine what a human would do in any situation. The next generation will most likely learn how to think and speak without ever seeing a word of human generated text. Now that LLMs exist to evaluate their output we have the necessary building blocks to design the generation of AI that can really produce content far better than any human.
We are living in the era of early-to-mid 90s chess engines, where AI learned from human moves and brute forced its way to barely beating the best in the world. It took 10 years from the point where chess engines could compete with top humans players until chess engines were effectively unbeatable. And those unbeatable engines were trained by other AI, not by looking at human games. It got to the point where looking at how humans played would have just made it worse.
That's why they compared it with YouTube shorts, which is basically YouTube's TikTok replica.
Also funny that the alpha males always go straight to the police and government when they feel harmed or at risk. And they fall over themselves to give government and police limitless power at the slightest provocation. Curious!
I'm not sure if that's true, but it would make sense if it was. Alpha males (ignoring there is no such thing) are the most restricted by laws, because they are the most likely to get their way if there was no enforcement of laws (through the natural law of might makes right). So it shouldn't be surprising that they'd be the least likely to let something slide without getting the police involved to get the same outcome they could have done themselves if the laws allowed them to use their own physical force. "Less alpha" individuals may be more used to not getting their way regardless of the laws, are therefore feel less entitled to get the police involved as quickly.
Indeed. I pay $20 per month for Cursor, and it works great. Why should I pay 15 times that much to be Elon's beta tester?
My guess is you don't use Cursor very much if a $20 monthly subscription is enough for your needs. I still run into limits with Anthropic's $200 per month (I have 13% of my weekly allotment left with 13 hours left in the week), and I just use it as a hobbyist. Even though I am a very active hobbyist, I still can't imagine someone using these tools even 10 hours a week on just a $20 per month plan.
ChatGPT was trained on all the stuff that happened in the past, and none of the stuff that will happen in the future.
Who do you know that was trained on stuff that will happen in the future?
It is insane that the EU hasn't done more to create local tech companies to reduce their reliance on the US. They need their own version of Baidu, Alibaba, and Tencent (among others), just like China does. It's fine to leverage allies for certain parts of your economy, but the tech sector is right up their with military when it comes to industries where the EU shouldn't be depending on external allies so strongly. It's not like the EU has the same religious devotion to free markets that the US has which would make them hesitant to prop up their local tech companies for 10+ years until they could survive on their own.
I found an EU report from 2025 that suggested it would take $5 trillion to do this, which would be about 5% of the government revenue of all EU countries combined if done over a decade. Just like efforts to become less reliant on the US military complex, the EU should really get started.
I wonder what they will do when the cost of AI increases? We all know that AI companies are selling their services at a loss. Often on a cost-of-compute- basis, but even more so when you factor in model training costs incurred with investor cash. And that is even before we account for how the shortages of relevant hardware and server space for running all of this are driving up the costs of memory, chips, etc. Or the fact that the energy crisis is only getting started, and will impact literally every part of the value chain for addressing the current and future demand.
In 1998, 1 Mbps of bandwidth cost $1200 per month. Today it is about 10 cents. The past never perfectly predicts the future, but I wouldn't be on the side who thinks AI won't be significantly cheaper in the near future. My Claude subscription costs be $200 per month today and gives me $2500-3000 worth of tokens per month. But in 5 years that same amount of usage will probably be a few hundred, and in ten years it will probably be $50.
The quality of
You obviously weren't around for the petrified with hot grits era of Slashdot.
Are you implying that parents are more qualified to determine what's best for their children than the government? Keep talking like that, and you'll end up in a reeducation camp.
I'd say the government is far more capable of determining what's in the best interest of their children than the parents, but in its current state they don't leverage that capability or even have a desire to do so.
The chances that a parent has the same access to child psychologists, researchers, teacher's associations, and any other groups necessary to determine the child's best interests is laughable. The chances that a parent will base their decisions more from their own biases and ignorance than on careful research is high.
But the chances that the government in practice will do all of those things, and put in practices to effectively allow parents / teachers / etc to provide feedback on an individual child to make exceptions to broad rules, are also laughably low. So in practice it's far better to allow parents to make those decisions, even though I wouldn't consider it ideal.
Agreed completely. If a parent helps the kid register, there shouldn't be any problem here. Working as intended as far as I'm concerned.
Why not just use Unreal Engine? I'd do the same thing as this parent.
Problem is, if you repeat the question the LLM will give a different answer each time.
No it won't. It may change the wording, but not the answer. I just asked Gemini what the capital of the US was five times using Google, and got five unique responses. All of them said it was Washington DC though.
Google has implemented Trump Mode in their AI?
No, they said Google tells the truth 90% of the time, not 10%.
NOWPRINT. NOWPRINT. Clemclone, back to the shadows again. - The Firesign Theater
