I've got a couple more contracts for my day job I'm finishing up and then I think I'll have the cash to build the demonstrator. (Even cheap satellites cost alot, when one offs) Design is still ongoing. Oh, and an announcement after those contracts finish up. Stay tuned!

@melqtx I don't know if the profile picture or the claim that Java will replace Python is funnier. Honestly, both have their place. Python is for those that don't want to write boilerplate code to make a string for AI work, and Java is for those upgrading from Cobol enterprise stuff.

@bubblebabyboi It's a topology joke. By the rules of topology, these are all equivalent to the object they're labeled with. If you can morph one thing to another without cutting/pinching/etc. than those two things are topologically analogous. Turns out the core differentiator is # of holes.

I got bit again on cache invalidation. If you can't seem to install numpy on Blender's local Python instance no matter what you try, don't forget to run the pip reinstall with '--no-cache-dir' and '--force-reinstall'.

@yacineMTB @nearcyan I've had a few instances over the last few days where the LLMs didn't help but old forum posts and the like did. I'm afraid to say that it's very much still alive.

@techhsiren Ok, what I hear is: step 1. find gal, step 2. marry, step 3. make cloths, step 4. she wears?

@mov_axbx @EricHallahan Ya, jumping right in is a great way to do it. The entire thing is sub-10k of code, and you don't have to understand all of it to get up and running on it. Fun project. I commit to it every once in a while when I'm angry enough at Nvidia to do so.

@mov_axbx ooo, do let us know if you get it. I'd love to see it.

I'm being a bit coy about the details, but the core of my observation has to do with the fact that, if you look at the Big-O notation of the evolutionary algorithm, the actual algorithm would fall under a much more favorable Big-O vs what a random algorithm would.

@qtnx_ I've poked at this a few times and came up with a few basic points: 1st, if you scale a Turing machine enough you can get human brain level abilities. We've know this for over 100 years. 2nd, the BNN equivalent of scale is scaling Brodmann areas 22,44,45 to whole brain level.

@lauriewired I learned my lesson already, but, for the record, I retaliate:

@AdrianDittmann Among other things, a beautiful demonstration that 'equals' is not the same thing as 'assigned to'.

@akapoor_av8r Considering how we're straight up doing RL training mostly in simulation these days, it would track that the hardware requirements become less and less. Also:

@codyaims Ducted props for the win. You gotta love that extra lift.

@TheCatholicEngr "In the year 2054, the entire defense budget will purchase just one aircraft. This aircraft will have to be shared by the Air Force and Navy 3-1/2 days each per week except for leap year, when it will be made available to the Marines for the extra day." - Norman R. Augustine

Ok, it turns out that trying to simulate an entire country is...memory intensive.

@basedpm I'm trying to take cross-sections of OSM2World 3d models and use them to simulate environmental conditions for a specific day job project. I'll be able to reveal more about what that is fairly soon, but you'll have to wait a little while longer before that's the case. Soon. ;)

@HaruhiAisaka @ChristianAniPod You know, given the selection, he tried his best.

@djcows What a whiplash effect from his, reasonable, demand that you read "How To Ask Questions The Smart Way" (Eric S. Raymond) before asking questions on spaces he controls. A person needs a high throughput low latency good sig./noise question, an AI does not. Saw it coming, love it.