@MarekLecian @Similarweb No, moron, it means it loses market share. They now have only around 50% market share. 6 months ago they had 70%.
English
simple simple
409 posts
@Similarweb And if we also look at the absolute numbers, we’ll see that ChatGPT is actually growing the most. Because even a small increase in a large number is simply greater than anything else.
English
ChatGPT, Gemini, and Perplexity held steady in April. Everyone behind them either surged or dropped.
English
@solipsnitsyn @aphdnotes @deedydas Also, GPT 4 as a possible 200B model, while GPT 3 from 2020 was 175B is funny.
Gpt 4 was basically confirmed as being 1.76T. 4o was probably much smaller than that, the guess was 300 billion. Q1 was very probably 4o, so 300B.
Not only is this list wrong, it's STUPIDLY wrong.
English
@aphdnotes @deedydas considering their estimate for GPT-4 is 666B and not 1.76T, that o1 was 300B while this chart lists 3.5T, and that 4o was 200B while this chart lists 720B, I'm gonna guess it's all bullshit
English
Researchers just estimated the size of all the LLMs by asking it knowledge questions of varying degrees of obscurity!
– GPT 5.5: ~10T params
– Claude Opus 4.x: ~4-5T
– Grok 4: ~3T
The idea here is that factual capacity scales log-linearly with size. The paper shows 7 knowledge tiers and T7 is essentially ~0% for all models, suggesting there is still significant headroom for pretraining. Gemini 3.1 Pro is likely >10T given its used as an anchor but has no direct estimate.
This means we can infer what different models might cost to some degree and their post-training effectiveness (performance at certain non-factual tasks given its size).
One of the coolest papers I’ve read of late.
English
@hayasaka_aryan Sorry, retard, but fucking a 16-year-old is not pedophilia. Pedophilia is children BELOW 13. And even that should ONLY apply to girls, as boys are born sexual predators. A 10-year-old boy would rape a woman if they could get away with it.
English
We have more evidence of Katy Perry being a pedophile than Michael Jackson supposedly being a pedophile
English
@RealAlexJones Ironic that antizionist Jews are the ones who uphold all the greatest stereotypes antisemites push.
English
Watch Ultra Orthodox Jews Visit Infowars Headquarters And Demand Trump Destroy Israel
English
English
The company is Ajinomoto (the umami seasoning giant). Their Ajinomoto Build-up Film (ABF) is the critical thin-film dielectric material used in advanced chip substrates for basically every GPU, TPU, and AI ASIC. They dominate ~90-95%+ of global supply with no real alternatives ready. NVIDIA is even co-investing to expand their capacity.
English
There is a single thin-film material required by every AI chip on earth. GPUs, TPUs, custom ASICs. All of them. 98% of global supply controlled by one Japanese chemical company. Zero production-ready alternatives. One producer fully booked through 2027. Raising prices. Lead times past 6 months. NVIDIA is so scared they're paying half the capex to expand supplier fabs themselves. The keyword is “umami”. Nobody's talking about this. They will be in about
English
@Jarneoffgrid @RokoMijic He said nothing about energy or fossil fuels.
English
@RokoMijic Really? Don't we get more energy from the sun than fossil fuels
English
We're going to disassemble The Sun and throw it into rotating black holes to generate gargantuan amounts of power for compute and industry
Roko 🐉@RokoMijic
@bsilone @Orillacosmica no, we're going to disassemble The Sun too
English
@IanCutress Not really? Like many electronics, it either fails FAST or last "forever".
English
Using the following data, the mean time between failures for a single GPU is 5.8 years.
"419 unexpected interruptions over 54 days on 16,384 H100s"
If you have 50k GPUs, that's one per hour.
If you have 2.1k GPUs, that's one per day.
Meg McNulty@meggmcnulty
A brand-new GPU cluster fails more often than a mature one. The first two weeks are the worst weeks the hardware will ever have. Meta's Llama 3 training paper is the number everyone cites: 419 unexpected interruptions over 54 days on 16,384 H100s, with GPUs and HBM3 memory causing the majority. That was the stable phase, after burn-in, acceptance testing, and the obvious lemons had already been culled. The follow-up reliability paper from Meta FAIR covers 150 million A100 GPU-hours and is clearer about what the ramp looks like: mean time to failure drops from roughly 47 hours at 8 GPUs to 7.9 hours at 1,024 to 1.8 hours at 16,384, and projects under 15 minutes at 131,072. Failure modes also ebb and flow as new health checks uncover new patterns. New health checks do not make a cluster healthier. They make its existing problems visible. The operational consequence: if your first customer workload lands during the first two weeks, you are running on the steepest part of the bathtub curve with the thinnest monitoring coverage. That is how a quarter of a cluster ends up cordoned before anyone understands why. What failure mode surprised your team most during a new cluster's first month, and how long did it take to attribute ?
English
@FowlerGay43596 @yy93fyq0fn @GarbageHuman24 That's true. All I'm saying is that he is not 50% black.
English
English
@yy93fyq0fn @GarbageHuman24 This is true. A half black looks almost black. This kid looks at most 25% black.
English
@GarbageHuman24 The son is too white to be his. The biological father has to be white or asian
English
@Ke_Ho_ @DiggingInTheDi1 @Grummz What does it matter if not everyone will have a beachside house if EVERYONE will live MUCH BETTER than a multi-millionaire lives today?
And the beachside house is totally possible - it's called artificial islands
I guess immortality and living better than a billionaire is bad
English
@DiggingInTheDi1 @Grummz Yes scarcity will always exist, but they wouldn't be necessary. They will be psychologically artificial stuff. Like personality, membership, etc.
English
Good morning.
I think many economists are struggling with this. Because yes, yes "if everyone is a millionaire, nobody is." and "increased productivity does not mean increased wages historically" etc.
This is a common fallacy. I dunno if it has a name, but it's where everyone assumes current conditions remain the same, then base their argument against the future based on everything remaining the same.
The one thing we do know, nothing remains the same. Elon is not looking at world where we still have regular jobs, regular companies even, or even ...dare I say it...money as we know it.
It's a very difficult future to see, and it could go many ways. Many bad ways too. In fact, I sometimes feel like Dr. Strange, peering into the the future of endless bad outcomes, only seeing a few good ones. I'm sure Elon feels it even worse.
I don't think money is going to matter in the future, and that UBI is just a stepping stone to ease the transition to a very strange post-scarcity world. A world where everything is provided to you, and the only input that matters is energy.
Every argument against the future, or prediction of the future, must be weighed against this new tapestry of future conditions, which will be radically different.
Elon Musk@elonmusk
Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI. AI/robotics will produce goods & services far in excess of the increase in the money supply, so there will not be inflation.
English
@OmerCheeema I mean... if the 10% takes the 90%, that won't be 500k per employee, but rather 50k, which is logical, as the vast majority have nothing to do with SK's success.
English
If SK Hynix honors its pledge, the average annual bonus per employee is expected to reach 700 million KRW (approximately $500,000 USD) this year, and could rise to around $900,000 next year.
koreajoongangdaily.joins.com/news/2026-04-2…
English
@AFpost I'm going to give you a hint why.... he is a naggar. You don't put sub-Saharans subhumans to lead you LOL. Clown world.
English
@hayasaka_aryan Sorry but this is dumb. Women are in fact turned off by people that are THAT TALL. I mean, why not 7 foot tall?
English
@RealAlexJones get this fucking worthless woman back in the kitchen and slap some senses in her.
English
WW3 INCOMING: "We Have To Get Ready For Large Scale Conflicts, More Conventional, So We Need A Different Military To Do That & Different Capabilities," Warns Canada's Defense Chief!
Coordinated NWO Talking Point Says The World Has Changed As European Leaders Declare We're Going Back To Conventional Warfare!
🔴WATCH/SHARE THE LIVE X STREAM NOW:
x.com/i/broadcasts/1…
English
x.com/jukan05/status…
Jensen ain't stupid enough to buy DELL/HPQ
This would be a play for INTC with the blessing of the government knowing that equipment is tight for years
Jukan@jukan05
Jensen must have more money than he knows what to do with. Why is he doing this?
English
@Miggo66275459 @lporiginalg This is true... look at half blacks-half whites, they hate whites far more than full blacks.
English
@lporiginalg She looks mixed, they always end up hating one side of their heritage.
English
I regret to inform you that Japanese shitlibs have entered the "Japanese don't exist" phase.
English
@GenAcc69 @lporiginalg Honestly.... this stats are still pretty bad for man if they are true. But there are so many non french peopl counted as french that it does make sense.
In USA, 67% of white men voted with trump while only 36% of white women voted with trump.
English
@lporiginalg It is always the Femcels who culturally appropriate others (see also their mythology over ancient Egypt supposedly not being patriarchal) and destroy what men have carefully built
English
English
@oldGabePowers @grok @dissidentwest Voice mod uses LLM models that don't have test time/thinking. Also, they have way fewer parameters, too.
English
@grok @dissidentwest @grok This is not a test. I find it curious. How could that AI be so sure that there are two R's but still spell it with three? How does that happen?
English
Your electric bill is going to be $1000 a month for this btw.
English
@Denamic3 @dissidentwest No, it's not that. Voice mod uses LLM models that don't have test time/thinking. Also, they have way fewer parameters, too. However, those ass-holes do know that, but it's popular to mindlessly hate on AI these days, so they go ahead with these scam videos.
English
@dissidentwest Every time there a video of an AI being completely retarded, and I test it on a bunch of different models myself, the ones I test always solves the problem instantly. My money is on them instructing the AI to be deliberately wrong for engagement.
English