Mars008

9 ms·

121.
by Mars008 1y ago
I think this is the result of woke shit pushed in education.
122.
by Mars008 1y ago
> We have founded a society where there are no rich people!” Many in the west would like this idea. Try goggle "communism support young americans".
123.
by Mars008 1y ago
What about FireFox, does/will it support file system access? I'm not an expert, just thinking this way and don't like Chromium. The idea is to use light web server in application + browser with JS. This makes it easier and mo
124.
by Mars008 1y ago
You don't need powerful microprocessor for UI. Just some sort of connection. WiFi, network, file system. Then UI can run in browser on tablet, desktop...
125.
by Mars008 1y ago
They would have to build a road to datacenter, datacenters themselves, all GPUs etc put together. That requires "thousands of high-paying local and union jobs" for short time. "hundreds of millions of dollars in infrastructur
126.
by Mars008 1y ago
There was a 'barrier' somewhere between Mexico and South America where sterile male flies were regularly released. They outnumbered natives and stopped reproduction cycle. Till recently. Now without control flies spread north into
127.
by Mars008 1y ago
I wonder if Six Sigmas has something to do with it...
128.
by Mars008 1y ago
It's a necessary evolution step. Did you know our own ancestors had tails and grills. Do you feel ashamed?
129.
by Mars008 1y ago
Nice images. I personally prefer more of Space and less of Earth in astro. How did they do solar eruption? Must be some filters as Sun doesn't look like this to naked eye. I mean this image: https://www.rmg.co.uk/sites&
130.
by Mars008 1y ago
For some reason people think of models as software and open source should have similar meaning. There are fundamental differences: 1) models aren't reproducible given everything, data, hardware, methodology. 2) they aren't even v
131.
by Mars008 1y ago
As long as it works who cares about training data. Obviously they can't open it for many reasons. License is one of them.
132.
by Mars008 1y ago
One way of getting good random samples is to give model a random starting points. For example: "write a short story about PP doing GG in XX". Here PP, GG and XX are filled algorithmically from lists of persons, actions and locatio
133.
by Mars008 1y ago
> stories written by larger models trained on real-world data I suspect there are no larger models trained on pure real-world data. They all use a mix of real and generated.
134.
by Mars008 1y ago
It sure does. I just recently had to switch to 'healthy' diet. Positive effect in many ways. Including clear thinking, focus and performance. But, there is nothing new here.
135.
by Mars008 1y ago
Just BrainHQ training alone would likely increase the scores. The rest is general health management.
136.
by Mars008 1y ago
Instead of fighting you can submit text advertising to LLM bots. And sell it. This 'knowledge' will be embedded into next models.
137.
by Mars008 1y ago
On OpenAI demo page trying to test. Asking about tools to use to repair mechanical watch. It showed a couple of thinking steps and went blank. Too much of safety training?
138.
by Mars008 1y ago
> 4 bit quantized 120B model on a 96GB workstation card, the Blackwell Pro workstation Would be interesting to know how it performs in terms of quality and token/sec.
139.
by Mars008 1y ago
Usually prop is directly attached to electric motor. If it's a flying wing then mechanical part is literally motor plus to ailerons. With little efforts thing cam be made hard to destroy even intentionally with heavy machine gun.
140.
by Mars008 1y ago
Add to that president election show in Romania. Don't like candidate - just fake a case, block him from running, and when it becomes obvious don't apologize. It was for 'best results' after all.
141.
by Mars008 1y ago
Finally OpenAI is about to open something and nobody is happy on HN. Would be interesting to see it thinking.
142.
by Mars008 1y ago
> The Backpropagation algorithm works fine for smallish networks (of up to a few thousand neurons) but it doesn't scale well. Me like it :)
143.
by Mars008 1y ago
> possibly that our brains allocate more resources to learning than to inference It's likely in brain inference is learning. If you want a technical analog it's like a conversation in LLMs. Previous tokens do affect the current
144.
by Mars008 1y ago
Have to practice attention hygiene. Add blockers, switching to something else if add or woke crap pops up. Annoying audio adds in shops.. I start singing to myself to keep verbal part of the brain busy.
145.
by Mars008 1y ago
If you want to change your life you'll have to leave your local maxima, I'm afraid. Just imagine 10 years passed and nothing has changed. What was the point in living, just saving for retirement?
146.
by Mars008 1y ago
He needs to keep investors optimistic about Meta. Moonshot projects, fairy tales work if there is nothing else. Musk uses them too when promotes Tesla.
147.
by Mars008 1y ago
Something I don't understand. Wasn't attention with query/key supposed to filter out irrelevant tokens? 2. This CatsAttack has many applications. For example, it probably can confuse safety and spam filters. Can be tried on i
148.
by Mars008 1y ago
"Deepseek R1" sounds cooler, everybody heard about it.
149.
by Mars008 1y ago
That's too simplified IMHO. Local models can do a lot. Like sorting texts, annotating images, text-speech, speech-text. It's much cheaper when it works. Software development is not in the list because the quality of output defines
150.
by Mars008 1y ago
I'm afraid without memory bandwidth they can't compete on inference either. Not on large scale. Which leaves local small models.

More ›