I took a video I shot back during Covid, stripped the audio, and threw it into a quickly composed GNU Radio flowgraph, and it sure looks like there are a few doppler curves in it, and you can almost hear a propeller plane around 150 seconds in... I live near a small airport, it was a nice sunny day, so someone surely was out flying.
Manual computational photography. Use any random digital camera with a deep field of focus, take many photos from slightly different viewpoints, then use Hugin (the panorama stitcher) to manually align on items in the desired focal plane. Then use it to project all the images to the same virtual focal plane. Then a little python script to average the images into the output. Crop and see if there are adjustments needed, repeat.
The images that focus on the Wrigley Building or the Tribune Tower in Chicago have a virtual aperture of about 10 feet across. For most subjects it's more like 6-12 inches in diameter.
I used to do the manual picking of point pairs on the train while commuting to Chicago. It was quite relaxing.
I think the "electromigration" part of this is a red herring. It seems perfectly plausible after reading this report that improper construction and low safety factors were the cause of the collapse.
Given the choice between "you should've gone up a size" and "you should've bought more engineering" these reports will always say the latter because they are prepared by engineers.
Yes, side channel attacks are sneaky, and always end up leveraging someone's hubris, with huge effort.
If you rely on an encryption machine that can't give the same letter on the output as the input, and the next thing you know, after millions of zł and £ and $ of effort, including capturing a submarine in the middle of the ocean before it sinks, others are reading your encrypted messages.
You isolate your Uranium enrichment facility in the middle of Iraq with guards, barbed wire, and pretty good security, but someone is able to slink in, put a passive monitoring probe on the 3 phase power going into your facility, and sneak it back out a few months later... and then they do an FFT, then measure the peaks to back-compute the speed of each stage's centrifuge, and the next thing you know, they can tell exactly how well things are going for you (read about that on Slashdot more than a decade ago! Possibly related[1])
If you put all your data through a digital computer, someone may use Van Eck phreaking to spy on your video output, or other emissions. If they have the right gear in range. Getting that gear in range again usually entails the results of large sums of effort.
Sure, it's possible to leak info out the data lines going from a CPU to RAM with a properly formatted set of patterns. If you don't even try to isolate that CPU from an ingress of control, however, you don't need that high effort strategy, you just need a good North Korean hacker working from the other side of the globe.
Folks, we mostly solved the ingress of control problem back in the 1970s, and it was getting folded into Multics as multilevel, capability based, security. But, the folks at Bell Labs, and many other places (MOST other places, and almost all of us, myself included) decided that security wasn't really important. That's why we're in this mess today, not because of idiots with LLMs and too much money.
We're in this because Operating Systems that rely on ambient authority can not ever be made secure.
Unix, Linux, Windows, and all the rest are unfit for purpose in the era of persistent Internet. We need to undo that local minima decision back from the era of paying by the minute for dialup connections to the BBS, and before every computer was networked. It's going to suck, but it's likely to be a Y2K level problem, not impossible.
I believe that AI is being massively underpriced, in an economically unsustainable manner. There will be less of it, but the hardware to run it locally will get much cheaper.
I believe that AI is going to take some jobs, just like automation of data handling means we no longer have typists, file clerks, etc.
I believe that AI should stand for alien intelligence, as it's essentially a random set of weights that's been trained on a trillion bytes of text, for a trillion passes. Think of it as speed-running evolution. Each new AI should be seen as coming from a different star system.
I believe that we can radically reduce the power requirement for AI with a different computing architecture that removes the distinction between RAM and CPU.
I don’t agree that it’s being sold below cost, at least as far as inference is concerned. One piece of data on which we could base our assessment, for example, is the DeepInfer release that DeepSeek published last week, in which it also stated that a cluster generates around $38 per GPU per hour, even at the lowest rate. In 2025, he had estimated the cost of renting a GPU at around two dollars per hour, with energy costs amounting to around 0.10 per hour per GPU. Obviously, this cannot be applied directly to Anthropic and OpenAI, as they publish figures for their own models and are likely larger organisations. However, even with costs that are many, many times higher, the margin on a single token remains substantial. It must be said that the price covers not only inference, but also training, research, free users and investments that need to be recouped. So ‘the individual token is profitable’ and ‘the business might not be sustainable’ could both be true. The separation between memory and processor isn’t going to disappear any time soon, but unified memory is already available in laptops, with an RTX Spark costing between $3,000 and $4,000 for a usable configuration. With competition – for example, from Intel, which has been working on power efficiency for years – prices will fall and on-premises models will become the norm, particularly in businesses.
Having ended up in the middle of a 5 car sandwich going to last years VCFMW, I've learned we all collectively fail to understand that the car in front of us can be stopped, in addition to using it's own brakes.
AI is massively underpriced at the present time. That's what's driven adoption so hard, along with the hype.
Once that ends, the prices will rise, quite a bit, and this will cause demand destruction, possibly causing most?? of the big players to fail. The thing is, even if we end up awash in AI hardware and RAM, the power to run it is a major part of the cost, and that's not going to get cheaper.
So, in a few years, there will be far, far less AI in use. We'll all be fine.
But what about technological advancement? Technological advancement is so fast that in 10 years, inference cost may drop heavily, and even deepseek 4.1 flash 1024 is very cheap and makes 80% in profit
Indeed. If current frontier models are already so disruptive now, we may have vastly better models in a few years but also the current models will have become significantly cheaper. Exponential growth (or shrinkage) is definitely at play here, and should not be underestimated.
I'm retired too, but I'm worried that if enough people end up unemployed that'll break social security, and the markets and everyone's retirement accounts.
I'm medically out of the labor market, if SS disappears, I will end up homeless.
The weird limit they put on Social security contributions is strangling the system.
Weird? That was intentional, it was just maintained for decades by the weird politicians your generation was obsessed with to destroy the system by the time you all aged out. Sorry to hear you're worried about it becoming insolvent. You should also be worried about the working population _not really caring about you_. We have kids now. Good luck.
At least if I could still work, there's a chance I could hold on. It sucks being medically frail.
And yes, I do care about society in general, and societal collapse.... what worries me is that the right seems bent on taking away the social safety net that's keeping me alive, and the left seems to be worried about not making waves.
Curated feeds are definitely the best.
[1] https://youtube.com/@impulsemanufacturinglabora3280/videos
reply