Amazon Prime cycles through slow pans of static images for content they are promoting. The image quality is high, but several of them had a distracting shimmer in high contrast areas. It might be that the image resolution was too high and they are subsampling without mip maps, but my bet would be on not doing gamma correct texture filtering. Always use sRGB(A) texture formats for images! Copy from RGB or YUV image formats if necessary, making sure to perfectly match the pixel centers. If you are just using textures like a pixel blitter with no stretching or subpixel movement you won’t notice a difference, but slow slides of an image will highlight the issue.

I appreciate how Grok doesn’t sugar coat corrections.

Companies selling the dream of autonomous household humanoid robots today would be better off embracing reality and selling “remote operated household help”. Have teams of employees running them 24/7, with the option to reduce their workload as autonomous behaviors become viable. That would be genuinely valuable for some people, although likely still a money sink even with low cost labor. It would be the ethical way to gather the desired training data.

When I started working in python, I got lazy with “single assignment”, and I need to nudge myself about it. You should strive to never reassign or update a variable outside of true iterative calculations in loops. Having all the intermediate calculations still available is helpful in the debugger, and it avoids problems where you move a block of code and it silently uses a version of the variable that wasn’t what it originally had. In C/C++, making almost every variable const at initialization is good practice. I wish it was the default, and mutable was a keyword.

The comments on software patents made me chuckle. After selling both Id Software and Oculus, my continued employment contracts included “will not participate in software patents” clauses, and yet I got asked to reconsider in both cases. It wasn’t a lot of pressure, so I don’t hold it against them; it is just “part of the game” that executives consider themselves to be playing. My views on the negative societal externalities of software patents have only hardened over the years. It rewards parasitism.

DGX Spark appears to be maxing out at only 100 watts power draw, less than half of the rated 240 watts, and it only seems to be delivering about half the quoted performance (assuming 1 PF sparse FP4 = 125 TF dense BF16) . It gets quite hot even at this level, and I saw a report of spontaneous rebooting on a long run, so was it de-rated before launch?

John Carmack explains how he applies Nassim Taleb's "anti-fragile" concept to his work, enjoying the thrill of new ideas while accepting that many won't succeed. Source: Deep Thoughts Engineering Speaker Series: John Carmack

A post I made 7 years ago on another forum that I think is worth sharing: There's so many lessons in thinking seriously about the technology of space settlement from the past human settlement. The most important one, though, is one I can barely put in to words. As recently as a generation ago, it was understood that engineering was a process of working "with the grain" of the materials, the technology, the site. When I first studied architecture, for example, one was taught to study the site, the trees that would be preserved, to think about the sun and the shadows, the slope, the drainage of the site. Woodworking, then as now, is something one has to pay close attention to the material and understand that this piece of wood is not quite like that one -- and a generation ago, wood was much more commonly an engineering material. Composites still convey something of that attitude. Now, we have so much power, so much machinery, that all too often those subtleties pass us by. Building sites are first scalloped by bulldozer until every site is just like every other. Margins in machine design are made high enough that only in the highest performing applications does one strive for elegance in structural design. Whenever, and wherever, our first settlements off the Earth take place, we will find ourselves once again challenged by Nature, and forced to use the least amount of energy that will get the job done. It's so easy to look back on the technologies of five or ten centuries back and think they have nothing to teach us. But that simply isn't true. Will we need to do the same things they did? Of course not. But our challenges will need to learn some of the same attitudes. The Thule culture of hunter-gatherers made the Arctic ice their highway. They did it by inventing an "in-situ refueling" transportation technology -- the dogsled. By domesticating *carnivores* as transportation instead of herbivores, they gained the capability in the Arctic to "refuel" their transportation by hunting, in a land where there were no plants to feed to the herbivores other cultures depended on. The Polynesian seafaring canoe, adapted again and again to the trees found on a new island, a sophisticated technology of wood-shaping tools, made from basalt rocks ground down, shells, polished with stones, drilled with pump drills, fibers from coconut made in to cord, sails, rigging. Techniques to bring not just food and water, but whole ecosystems -- live plants, seedling trees, domestic animals. Space settlement is going to require us to really rethink how we make things, what materials and techniques we use, until where we might look at the gray soil of Luna as a barren place, the trained eyes of a future generation will see the rocks and the regolith, the slopes and the rilles and the lava tubes, as rich resources.

Anyone here on the Amazon Kindle iOS team? For ages now, the iPad kindle app has locked up for many seconds at a time when downloading large media files. Maybe a second or two wasn’t considered a big deal for downloading a book you are going to read for ten hours, but when you buy a half dozen or more large file size comic books every week, it grates. Long ago, it didn’t behave this way, and you could interact freely while multiple files were downloading. Someone broke it.

I view the proliferation of Halloween skeleton sizes through first edition AD&D monster manual heights: 3’ -> kobold, 4’ -> goblin, 6’ -> orc, 10’ -> ogre, 12’ -> stone giant. Waiting for the competition between Lowe’s and Home Depot to bring a storm giant!

Has Maximum Power Point Tracking (MPPT) technology gotten dirt cheap enough that it is integrated into solar powered lawn lights? At some point, the tradeoff between “slightly bigger panel” and “added power electronics” crosses over.

Blindsight by Peter Watts (2006) is my standout fiction book of the year, by far. Grand, challenging, scary, and super autistic.

I helped fund the restoration of an old IBM S/34, and now I find myself occasionally thinking about what realtime games you could make run on a 5251 DisplayStation. Tetris would probably be the most engaging. Not sure if you could do it in RPG or COBOL; might require S/34 Assembly. gofundme.com/f/ibm-system-34… piped.video/@crusty.computer en.wikipedia.org/wiki/IBM_Sy…

celebrating 25 years of Quake 3 CPMA: the most skillful, hardest, fastest, greatest FPS of all time. the world's best esport. the true start of the aim community and freeform movement. the first frag movies and the birth of montages. watch it all at once. holy shit! by entik

I expect AI to produce a lot of value while trying to make a coherent body of knowledge out of its training data, but on its own, AI is limited to a sort of ancient Greek philosophical process, where the method is rhetoric, rather than the experimental process of science. This has always been one of my gripes with the AI doomer crowd – even infinite intelligence can’t derive all of physics, let alone biology or sociology, from first principles. In a finite, completely specified environment like Chess or Go, you can indeed keep diving deeper and deeper with nothing but computation. Relatively few endeavors outside of math and some parts of computer science fit that. Elsewhere, you can make a carefully considered hypothesis, but then you have to go and test it. If you ignore that, you are in the offline reinforcement learning regime, and run a very real risk of iteratively bootstrapping a highly confident fantasy that is far divorced from reality. Grok should write grant proposals for resolving important conflicts in The Knowledge.

Numpy’s fromfile()/tofile() make me smile — no typing or byte order metadata, they just write/read the raw bytes. A joyful API! I wish PyTorch had them natively on tensors, so I wasn’t doing so many tensor.cpu().numpy() operations.

Sometimes I see papers with hyperparameter sweeps over 0.001, 0.003, 0.006, 0.01, etc. Many hyperparameters are better expressed in negative integral log2. Small values like learning rates directly, and values close to 1 like EMA factors and TD lambda / gamma with 1-2**val. It is interesting how many parameters are relatively insensitive to doubling or halving, and need bigger changes to reliably move the results.

Teslas displaced my exotic cars many years ago, but with the Roadster still an indeterminate distance in the future, I had been feeling an urge for something a little more… analog. I was initially thinking about a 60’s Camaro with modernized running gear, but a part of me was also nostalgic for my old MGB. I eventually remembered that there was a car right at the intersection of muscle car grunt and British roadster charm – the ‘65 Shelby Cobra. It is kind of amazing that the federal Low Volume Motor Vehicle Manufacturers Act of 2015 authorizes small runs of replica vehicles to basically ignore all the safety and emissions regulations. All the paternalism and crusading was just pushed aside in the name of letting small companies make badass cars. On my birthday, Trista nudged me into visiting @EMotorcars where they had a dozen different replicas on hand. I was partial to the idea of a carbureted 427 with a magneto as a Mad Max post-EMP LARP, but practicality prevailed, and I wound up with a modern Coyote crate motor in a @Backdraftracing chassis.

Thank you so much for playing Walkabout Mini Golf over the past 5 years! We wanted to take a moment to look at some fascinating numbers and recap what our incredible players have been able to accomplish thus far! Read more here: mightycoconut.com/blog/5-yea…

The audience for this is small, but we have an open source repository for the “Physical Atari” work we did at Keen. Working purely in the physical world is a huge burden compared to simulation, but it is important to have a reasonable grasp of the gap between the two. The “robotroller” build has a separate page: robotroller.keenagi.com/ And the code to drive everything is here: github.com/Keen-Technologies…