Any cryptography you’re likely to encounter uses fixed size primes over a residue ring for performance reasons. These superlarge primes aren’t relevant for practical cryptography, they’re just fun.

Not bad, but you’re missing that the Bluetooth device can report audio latency back to the source so it can delay anything that needs to synchronize. In practice there’s half a dozen more buffers in between and a serious tradeoff between latency, noise sensitivity, and bandwidth.

Extradition treaties are almost always reciprocal and this particular treaty is publicly available. No public treaty is going to include a promise not to coup another government because of the obvious political consequences of admitting you might to everyone else.

No, the “non-fungibility” simply means that anyone who creates an NFT with the same link will be distinct from your link to the image, even if the actual URL is the same. Both NFTs can also be traced back to when they were created/minted because they’re on a blockchain, a property called provenance. If the authentic tokens came from a well known minting, you can establish that your token is “authentic” and the copy token is a recreation, even if the actual link (or other content) is completely identical.

Nothing about having the “authentic” token would give you actual legal rights though.

That’s perfectly solveable with math. Each grid square can take 10 colors, so there are 10^100 possibilities. That’s about 330 bits of entropy, or equivalent to a 51 character password. That’s gross overkill if the underlying cryptosystem isn’t broken, but insufficient if it is (depending on the details).

Cryptography routinely deals with much, much larger numbers than what you’re suggesting (e.g. any RSA key), and even those get broken occasionally.

No. Nvidia will be licensing the designs to mediatek, who will build out the ASIC/silicon in their scaler boards. That solves a few different issues. For one, no FPGAs involved = big cost savings. For another, mediatek can do much higher volume than Nvidia, which brings costs down. The licensing fee is also going to be significantly lower than the combined BOM cost + licensing fee they currently charge. I assume Nvidia will continue charging for certification, but that may lead to a situation where many displays are gsync compatible and simply don’t advertise it on the box except on high end SKUs.

Flat cables can be conformant and they still have twisted pairs. Cables just have to meet the physical properties set by the standard.

Other than Apple music and iCloud, they’re generally less intrusive about popups than Microsoft. Their tactic is to completely prevent competitors from integrating with the system at all rather than nag you to use a setting. For example, there’s no way to use Google maps or Spotify in all the same ways you can use Apple music or Maps.

A torque converter is part of the whole transmission system even if it’s a separate housing. When you buy a new transmission, it comes with a torque converter.

Torque converters also create the majority of heat in automatic transmissions and are why automatic transmissions get coolers in the first place. How many manuals have you seen with transmission coolers?

deleted by creator

There is independent government oversight. That’s NHTSA, the agency doing these investigations. The companies operating these vehicles also have insurance as a requirement of public operating permits (managed by the states). NHTSA also requires mandatory reporting of accidents involving these vehicles and has safety standards.

The only thing missing is the fee, and I’m not sure what purpose that’s supposed to serve. Regulators shouldn’t be directly paid by the organizations they’re regulating.

Just for context, a large chunk of “top tech talent” at the companies in the study are going to be making 200-400k. While there’s still going to be issues with pay, it’s a pretty different situation than fast food workers or similar.

I’m not assuming it’s going to fail, I’m just saying that the exponential gains seen in early computing are going to be much harder to come by because we’re not starting from the same grossly inefficient place.

As an FYI, most modern computers are modified Harvard architectures, not Von Neumann machines. There are other architectures being explored that are even more exotic, but I’m not aware of any that are massively better on the power side (vs simply being faster). The acceleration approaches that I’m aware of that are more (e.g. analog or optical accelerators) are also totally compatible with traditional Harvard/Von Neumann architectures.

ML is not an ENIAC situation. Computers got more efficient not by doing fewer operations, but by making what they were already doing much more efficient.

The basic operations underlying ML (e.g. matrix multiplication) are already some of the most heavily optimized things around. ML is inefficient because it needs to do a lot of that. The problem is very different.

I couldn’t find official dimensional accuracy specs for any formlabs machines except the 1, which lists 150um. Perhaps you’re talking about the 3, which has a specified minimum spot size of 85um according to this paper. Where did they claim micron dimensional accuracy?

Wavelength has a very direct impact on the resolution you can print because it’s an optical system. Under perfect conditions, it’ll be diffraction limited, which is typically anywhere from several hundred nm to tens of microns. That’s an ideal system though, you’re actually going to be getting a dimensional accuracy somewhat above that in practice, probably tens to hundreds of um.

TCP has been amended in backwards incompatible ways multiple times since 1993. See e.g. RFCs 5681, 2675, and 7323 as examples.

Plus, speaking TCP/IP isn’t enough to let you to use the web, which is what most people think of when you say “Internet”. That 1993 device is going to have trouble speaking HTTP/1.1 (or 1.0 if you’re brave) to load even the most basic websites and no, writing the requests by hand doesn’t count.

eGPUs aren’t supported on Apple silicon.

You have no idea how modern technology is produced. Any particular product is usually the result of dozens to thousands of iterations, some funded with public money and many not. Let’s take an example from your chart: DRAM. I actually don’t know when DARPA “developed” DRAM (since DARPA usually funds private companies to do development for them), but it must have been before 1970 when Intel designed the 1103 chip that got them started. Do you think that pre-1970s design is remotely similar to the DRAM operating on your device today? I’ll give you a hint: it’s not.

And no, modern device development does not consist of gluing a bunch of APIs together. Apple maintains its own compilers, languages, toolchains, runtimes, hardware, operating systems, debugging tools, and so on. Some of that code had distant origins in open source (e.g. webkit), but that’s vastly different than publicly funded and those components are usually very different today.

They’re failing to produce competitive modems because modern wireless is one of closest things humans have to straight up black magic. It’s extremely difficult to get right, especially as frequencies go up, SNR goes down, and we try to push things ever faster despite having effectively reached the Shannon limit ages ago.

You can, today, download an app and go ride in a self-driving car around multiple US cities. All of those cars use LIDARs. Sensor disagreement is not a major issue because sensor fusion is a very well-understood topic.