numba888

well, "RTX Pro Blackwell" desktop and workstation version looks to be the answer for midrange. Some should be available starting in April.

Edge inference most likely. Its FP4 performance is about 1/3 of 5090, power 170W for the whole thing. It can run big model or several small. Shifting balance to memory favors MoE. Would be nice to see FP32 numbers, they are used in training. My guess about 20 TFLOP, may be more, but 5090 is still times better.

> glad to hear about DGX Station

how much is it expected, my guess barely fits in 5 digits. Would be nice to have something in between Spark and Station. I.e. some desktop withing $20K.

what's the better alternative, DIY AMD? No, thanks. I'm interested in the result, not the process. In terms of robotic software there is just nobody close. Any attempt to port it to different hardware will be just a big pain and limited result at best. So NVidia looks to be the way to go till China catches up.

cool robot at the end ;)

It looks to be the whole line with different performance. But no dates and times

2 hours in, don't see anything about Digits, their 'personal supercomputer'. Did I miss something?

It doesn't mean it's wrong. Like partial derivatives are centuries old. But we are moving on. Just recently I've seen a book on image recognition. While recent it was focused on math and not a word about NN. That looked so outdated. Someone spent the whole carer on this way. Another example would be linguistic approach to text translation. 'Stupid' LLM does it way better today. I can only guess how pro linguists feel about it.

2000-2003, both are pre-historic. We have neural networks now to do things like upscaling and colorization.

> At almost three pounds (.4 kilograms), Mixodectes was quite large

Something is wrong here with the numbers..

I did it some time ago. It works at low levels. Doing it in hard cases is non-trivial but possible. It's difficult to monetize that's why nobody is doing it seriously.

That would require keeping a library of 'given facts'. Which would include reproducible copyrighted materials or their derivatives.

That would mean with current resources AI can get so much more intelligent than humans, right? Aren't you scared?

In some cases it doesn't recover even without physical or chemical damage. Psychiatric clinics are full of this stories.

> where every LLM hits some arbitrary amount of context that causes it to "lose focus" and develop a sort of LLM ADD.

Humans brains have the same problem. As any intelligence probably. Solution for this is structural thinking. One piece at a time, often top-down. Educated humans do it, LLM can be orchestrated to do it too. Effective context window will be limited even though some claim millions of tokens.

Great, but how do you imagine multimodal with text, video. Just 2 for simplicity, what will be in the training set. With text model tries to predict next, then more steps were added. But what to do with multimodal?

At some point it will to punch him is the face and call a*hole. It should be somewhere in the training set if it has real life data.

> What are they gonna do then when all the discussion boards where that data would originate are either gone or optimized into algorithmic metric farms like all the other social media sites?

As people are using AI more and more for coding and problems solving providing company can keep records and train on them. I.e. if person 1 solved the problem of doing 2 on product 3 then when 4 is trying to do the same it can be either already trained into the model or model can lookup similar problems and solutions. This way the knowledge isn't gone or isolated, it's being saved and reused. Ideally it requires permission from the user, but price cuts can be motivating. Like all main players today have free versions which can collect interaction data. With millions of users it's much more than online forums have.

you probably need robotic leg for walking. or better pony. but doing anything physically requires at least working torso.

and a buddy who can get drunk too.

Actually it's even more complicated. every human has 50% DNA from each of parents. And at the same time 60% DNA common with banana.

I hope they both die slow painful death. As they are looking now for the last fool who will pay for everything. Musk is not doubt involved. Hope Trump is smarter than that and will not funnel taxpayers money (my money!!) into crooks pockets.

> The USSR didn’t actually have economic equality.

More or less it had. Except for privileged commie bosses everyone poor. "Rich" had cars (!) Having noticeable business was prohibited. Selling own potatoes on flea market, or growing chicken and pigs was fine.

> The US has drastically lower tax rates for capital gains.

In Mass. it's 10% to begin with.

> I’m saying economic growth is slowed through the economic inefficiency resulting from unequal taxation.

Not sure about this. Do you have examples when equal taxation helps?

OpenAI is going to add it to Plus subscriptions. I.e. available for many at no additional cost. Likely with restrictions line N prompts/hour.

As for API price, when it matters businesses and people are willing to pay much more for just a bit better results. OpenAI doesn't take the other options away. So we don't lose anything.

> slew of strong non-reasoning models (Gemini 2.0 Flash, Grok 3, Sonnet 3.7)

Sonnet 3.7 is actually reasoning model.

> You have no clue about the collapse of the USSR.

I lived through it. We can talk a lot about, but the fact is equality doesn't work. Maximum it gets to is everyone equally poor. And as a result of it there are no startups. You need middle class for this and motivation. Government controlled and own monopolies is also the result of equality. Nobody else has the resources. So commies had to create the economy, few bigger things are easier to manage. Managers aren't interested in risk taking, it's not worth it. Finally everything stalled and fell far behind the rest of the world. There were bright spots, but that's it. Like North Korea today can build missiles and not much else.

Just look around. Probably only Cuba and North Korea are still in true socialism. In both there is a small super-rich class and the rest is, yes, equally poor. Super-rich are pretending to be poor too and invent theories to keep the rest under control. BTW, I'm not sure Cuba still has strong ideology today.

Can we summarize it that 'easy' or 'low skilled' jobs aren't payed better today comparing with the past? Not surprising taking into account demand and population. More worrying is that high-skilled jobs aren't rewording. As for equality.. welcome to USSR, you see how it ended. Now only a few Arabs states have greater inequality.

> If so, I'm not sure that I really want to be a scientist anymore

I'm not sure you have many interesting options.

RAG can be part of the solution which is unlikely that simple.

The problem is this can be anything, not just mods and cracks. That's why I keep separate laptop for banking. This may not help if hackers take over the router. But still better than nothing.