Mars008

This is not mutually exclusive. They have home made robots and let others sell robots on their website. The same way they want to use AI and have resources to make their own. One way to use is to drive those robots. Another to enhance their web site. Current version sucks. I recently return the item because their bot told it has functionality while in fact it didn't.

Is that important? One misstep. She broke her leg half a year before.

Do you know any unbiased source? Any of these: BBC, NY Time, WSJ, CNN, MSN, The Guardian, NewsMax, MSDN...

How about miniPC + USB x bay enclosure? I'm thinking about it. Have 4 Synology NAS mostly as long offline storage. No problems with them in this role so far.

Not anymore. It's very usable when JS is not needed. I've seen more than 5MB/s Tor downloads. The good, or bad, thing is you don't control the exit point. You can only change it by resetting the circuit. Some websites are sensitive to user's location.

Same here, I have noscript almost always on. The problem is some things don't work without JS. Google and Bing search, youtube, even duckduckgo in plane FireFox. The later works in Tor browser, that's what I'm usually using. I usually skip on most other things that require JS to drive blinking ads.

And some jobs, offshored or not, are just human frontend to datacenters.

> And now everyone's $2000 Orins will be stuck forever on Ubuntu 24.04

Wasn't it supposed to run different systems? Unlike older NV hardware.

You are forgetting Thor has CPU too. So far we haven't seen how good Ryzen CPU is at matrix operations. Which makes given numbers 4 and 3 questionable at best.

The best would be to run a real test. However for robotics it's obvious, Thor with camera interfaces is unbeatable. They are used for (stereo) vision and obstacle avoidance. Add to that software stack.

if you don't mind single cable out of your console then your wish can be implemented with a powerful as you need gaming machine in another room. Or in the same with some noise canceling cover and/or watercooling.

People got spoiled with free content. Some never in their life payed for a newspaper. The problem is subscribing to one or two sources is not enough. You will miss a lot. To understand what happens in Chine you better read Chinese, and not BBC's interpretation. Now that there are so many free translating tools it's easy. The same about every conflict. There are two sides at least. Your favorite source will show most likely only one. Test it on Ukraine or Gaza.

> 5: There is still considerable prejudice against using AI to make game assets. I think some people (myself included) are hesitant to release a game with lots of AI generated assets at the current moment, for fear of public backlash. Eventually that will calm down and it will become more socially acceptable to use AI to generate game assets.

This is mostly from artists themselves. Most people are ok with even fully generated content if it's fun and interesting. Anyway, solo developer cannot afford even single artist, musician, writer.

That's where LLM helps. I recently tried GPT5 for story telling. Gave it a single image (in a bar, women, man, and a gun) and asked for a short story. Then asked for the next part 6 times. Every time at the end for illustration. The result was consistent and readable. Images generated had even similar faces. Remember, that was a problem with earlier models. I'm sure this will be used fill Amazon's bookshelves.

Now imagine that this is being used to train LLMs. Then fed back to users.

> how an interpreter would have to proceed to execute Recto code, it would still parse it into an internal representation and then execute that, and this internal representation no longer uses the 2D layout of the code.

Adding parallel execution will make it true 2D at execution time. Even though it's sequential code. More over it can be logically, architecturally 2D while being 1D from interpreter's standpoint. If it uses cooperative multitasking with 1 running task at a time. Or simply 1 core processor.

There is another big and growing group: charlatans (influencers). People who don't know much but make bold statements, select 'proof' cases. Just to get attention. There are many of them on youtube. When you someone on thumbnail making faces this is most likely it.

The same way most humans do, from internet. Generated data can be the result of processing yesterday's new by old model. It can be multiplied, repeated from different angles. This will make it more likely to stick in new model. But the best way is to add latest data to some storage which can act as a long term memory. In this case even old model will look fresh and up to date. I'm sure we'll get it soon. RAG can be considered as a primitive form of it.

It's not that important now as AI took off the ground. New models can be trained completely on generated data. That will give them core abilities. Real world knowledge... whatever humans can get models can.

There can be only one monopoly in each domain by definition. In AI world it's more like several 'fortresses'. Together they ruin click economy. Which almost eliminated printed books and magazines. Well, attention is limited resource.

If you put a tax on it they, employers, will simply pay even less. There will be enough people to work even for just food and roof. With the hope to get green card in a few years and bring the family. I was working for years with no promotions so that shitty boss could take the credits. I quit right after getting GC.

In two years there will be probably no new 'autonomous' LLMs. They will be most likely integrated into 'products', trained and designed for this. We see the beginning of it today as agents and tools.

Not sure there is on formula. Because there are two different cases:

1) performance constrained. like NVidia Spark with 128GB or AGX with 64GB.

2) memory constrained. like consumers' GPUs.

In first case MoE is clear win. They fit and run faster. In second case dense models will produce better results. And if performance in token/sec is acceptable then they are better choice.

That can be said about any infrastructure element. AI becomes one of them along with roads, telephones, electricity, etc. Should rich sponsor all of it to make poor around the world competitive? There are charities for this. Actually it's more complicated. The idea of world wide taxes was around for years.

US doesn't want gov debt even more. (not under dems ruling like last 4 year)

Money don't grow on the trees. Someone has to work or take debt. The question is should rich countries sponsor poor? Probably yes. How much? Trump decided enough is enough. EU want US to keep sponsoring the war. You know, they want it to continue but don't want to pay for it. So they try to hijack US government again by attacking Trump personally.

Immigration can be solved by border control. That's what dems fought so hard against trying to get more voters in. Another reason: they control food distribution.

Is it bad? It was trained on synthetic data with emphasis on coding and scientific thinking. Good on my opinion, that's what it can be used for. Not as universal do it all model.

You call it fair? 32 / 5.1 > 6, it's takes 6 times more to compute each token. Put it other way, Qwen3 32B is 6 times slower than GPT OSS 120B.

For me it's just a beginning. I can do now so much more.

> So, if the study showed experienced developers had a decline in productivity,

You forgot to add: first time users, and within their comfort zone. Because it would be completely different result if they were experienced with AI or outside of their usual domain.

I tried digging with excavator, I became lazy and fat.

Looks like today most people don't trust both sides. Some choose to ride one wave or another.

I think this is the result of woke shit pushed in education.