biggest jump since the reasoning era
>>110287889I heard there's a model or something that will only use small part of the model at time making it possible to run a multi 100 billion Gb size model on a potato.
>>110287889Does this account for time between releases? Opus 5.5 was like 40 days ago. The models are rolling out much quicker now
>>110288041once they figure out how to burn Opus 5.5 into a tiny chip it's actually over
>>110287889>morning hours in indiagm sar
>>110288261I have a strange feeling that, in the next 15 years or so, the sentiment of this post will become a quaint relic, like that old usenet post that said that a super nintendo emulator is impossible to make.If you think things are bad now, just wait until quantum computers get good. People will be so impressed that they can play gta 6 or whatever at 5000x times the performance that they'll gladly give up their local computing. And that's when the true datacenter hell will begin.
>>110288290You can't play on a cloud computer without a computer.That isn't how quantum anything works
>>110288277AI will always be a thinking white man's tech, mr. pajeet.
>>110287889vibeGODS won
>>110288316Yes, I know what a thin client is, I'm not saying that people will hallucinate it in front of them, you're missing my point
>>110288290>>110288316>>110288336They have to get quantum physics right first, it might be a long way off yet unless AI can shorten the timeline.And if they do get it right in our life time, computer games are not going to be one of the most interesting aspects of quantum computing, humankind will change exponentially and absurd science fiction possibly becoming a reality.
>>110288290>>110288435what are we getting
>>110288439>claude, run each scenario in picrel and report back what will be most likely.
>>110288451Verdict: No single tile wins. The most likely outcome is a messy blend centered on the top row, column 2: corporate-dominated, nostalgia-styled, high-convenience life. Pockets of automated ruralism and train-heavy Art Deco cities grow in the places that can afford them, and climate-driven nomadism grows in the places that can’t.The chart’s biggest omission is the most historically reliable option, which isn’t on the grid: muddling through. That means incremental change, uneven across regions, with every era convinced it’s living in a dramatic tile when it’s mostly just Tuesday.If you want my personal pick for the best achievable tile, it’s fully automated luxury ruralism, provided we solve the “who owns the robots” question first.
>>110288439Definitely, 19A0's vapor-futurist corporatism.
>>110288462>fully automated luxury ruralismonly for those who can afford it or were born in an upper class environment.the rest of us will be neo-nomadic modular post-urbanists.
>>110288462>with every era convinced it’s living in a dramatic tile when it’s mostly just Tuesday.I feel like every AI model was trained on thousands of smug filler news articles and buzzfeed quizzes
>>110288439Astra thought for 10 minutes on xhigh, here's the summary:Humanity’s most likely future is a mixture of these tiles: powerful corporations and governments, increasing automation, more life spent online, and unequal access to comfort. If forced to choose one, “19A0s vapor-futurist corporatism” is the closest institutional match, with parts of automated ruralism and gamer technarchy developing alongside it.
>>110288439rolling for Dyson Sphere but expecting Mad Max
>>110287889What are those numbers? What does it mean?
>>110288435Intel is very close to commercialize quantum computing
>>110287889Are those self declared?
Can it make a profit now?
>>110288290>they can play gta 6 or whatever at 5000x times the performance>implying normies can even notice anything past 60 fpsquantum computing cannot ignore latency you retarded pajeet. we had cloud computing powerful enough to stream gaming services 15 years ago.it's simply physically not possible because nobody wants to play a game with a 50-200ms delay on top of their own TV's delay, adding another 30-100ms.plus now we also have to deal with LLM's actually one useful feature, being able to quickly parse a fuckton of data, which leads to them being able to find a lot of vulnerabilities. which could actually lead to hardware, in fact, becoming much more localized.
>>110288041All Mixture-of-Experts models are like this which is basically all of them (that we know about)
>>110289137only two weeks away, in fact
>>110288439Everyone alive today are long dead before any of those could manifest.
