numba888

11 ms·

91.
by numba888 2y ago
P/E ratio is better indicator. Price/Earnings. NVidia: 46, Microsoft: 35, Apple: 34, Amazon: 50. As you see NVidia doesn't stand out much, it's even lower than Amazon.
92.
by numba888 2y ago
Yes, it has at least two distinct parts: consciousness and sub. First is visible to 'us', it's inner monologue or vision, or other senses. But we don't 'know' what happens in subconsciousness. The answer just p
93.
by numba888 2y ago
Many interesting projects, cool. I'm waiting to LLMs in games. That would make them much more fun. Any time now...
94.
by numba888 2y ago
> but an almost-memorized computation or proof is likely to be plain wrong hard to tell. never seen anyone trying it. model may almost-memorize and then fill the gaps at inference time as it's still doing some 'thinking'.
95.
by numba888 2y ago
I have different experience. It took me some time to make a rote and write down all turns. Now getting from location A to B is a lot easier. Take a look at proposed rote and make some corrections. Meanwhile I spend time thinking about somet
96.
by numba888 2y ago
you are mostly right. but seeing almost perfectly reconstructed images from training set it's obvious model -can- memorize samples. in this case it would reproduce the answers too close to the original to be just 'accidental'
97.
by numba888 2y ago
> you're liable to provide whatever updates/support the customer may need over the course of the software's licensing period to enable the customer to keep running it Does this include new layers for games, so that custom
98.
by numba888 2y ago
if they used it in training it should be 100% hit. most likely they used it to verify and tune parameters.
99.
by numba888 2y ago
You can try using online services with best models. With prices like $1 for 10K requests it's hard to make local competitive solution. I'm going to use this approach in my personal projects.
100.
by numba888 2y ago
Didn't expect it to disintegrate completely. At least they figured out what happened. Well, one step at a time. But that means no orbital flight for next couple of missions.
101.
by numba888 2y ago
Booster works, we've seen that before. No satellites deployment, no new heat shield test. Separation works. But that's it. Now the 'funniest' thing, this piece falls back where the ships are waiting. I hope it will miss
102.
by numba888 2y ago
they didn't get the telemetry after, what it was, 16 min(?) hope they'll find the reason which will be hard without black boxes like on airplanes. as every engineer knows it works flawlessly only at the end. if ever. the booster w
103.
by numba888 2y ago
4M viewers. comparable to top politics events. ship looks to be lost. this was the main part, so it's almost complete failure.
104.
by numba888 2y ago
look similar. the same person milking google for ads money? must be another service providing fake 'watches' and 'subscriptions'.
105.
by numba888 2y ago
> Not sure if what I said makes sense. Not sure either. But here is the lesson from this and other sources. To improve the output use multistep approach. Get the first answer, one or more, and pass it through the second verification step
106.
by numba888 2y ago
they don't self-host the models, neither embedding nor last step llm. taking into account low load self-hosting likely would be more expensive. if so why not to use the best models.
107.
by numba888 2y ago
yes, that's the whole idea, they want you to rent.
108.
by numba888 2y ago
1 working wheel is enough for balancing. they can be as simple as just a motor in rubber. it's more likely tail or wing will fail than both motors BTW, here is how it works from Boston Dynamics, it can jump: https://www.yout
109.
by numba888 2y ago
You can try to ask it to list all churches and assign them incremental number starting with 1. then print the last number. It's a variation of counting 'r' in 'raspberry' which works better than simple direct questi
110.
by numba888 2y ago
> reasonably flat comparing to wheel size. one can use bigger light wheels. that would make landing on short runways possible. besides, wheeling is much easier than walking. two wheels balancing and rolling around is not a problem today.
111.
by numba888 2y ago
it's just a generic idea. if you want to do something try to not to force yourself. will power is a limited resource. find an easy way. 'lazy break' is sort of application of this idea. the task may be complex, just mentally
112.
by numba888 2y ago
> but I cannot sustain that level of concentration for long periods You can fix this by paying attention to it. When your mind starts wondering acknowledge this, take 5 min 'break' lazily thinking about the task. Then get back
113.
by numba888 2y ago
There is correlation with race and groups within the same race. Which makes it more complicated than just one number fit all. I personally use my own self 10 years back as a 'normal' state. No I'm about 10% heavier then I sho
114.
by numba888 2y ago
> I wonder if my approach could be improved. Yes. When you use N batches by the number of cores the total time is defined by the slowest batch. At the end it will be just one job running. If you make batches smaller, like 1000/n_co
115.
by numba888 2y ago
This looks like a bigger brother of Orin AGX, which has 64GB of RAM and runs smaller LLMs. The question will be power and performance vs 5090. We know price is 1.5x
116.
by numba888 2y ago
> I sure wished I held some Nvidia stocks, they seem to be doing everything right in the last few years! They propelled on unexpected LLM boom. But plan 'A' was robotics in which NVidia invested a lot for decades. I think their
117.
by numba888 2y ago
Technically there is average IQ by country site, just google. Not that difficult to get faces by country. Put them together. Of course there are regulations and ethic. But in some cases it should work well and is more or less acceptable. Li
118.
by numba888 2y ago
Supper or not, it's way better for AI load then 4090 or 5090 set. And cheaper. I'm getting one instead of 5090 I wanted.