Any language designed for LLMs will probably be AST based. There is no need to convert tokens to text then back to tokens. Bu since LLMs don't understand code, no one will be able to see the bugs it generates.

I have a phone that hasn't gotten a security update in four years - it's patch level is February 2022, yet "Tap to pay" works. Google's attestation is not in any way related to security, but to control.

Windows 11 Notepad is a complete from scratch re-implementation using modern UWP components and packages as MSIX so it's not really the same app.

Not only that, the advertising enabled convenient excuses for layoffs that were caused by the downturn in economy (directly caused by higher interest rates and tax deduction amortization for R&D that took effect in 2022).

This is not about apps on the Play Store. This is about sideloading.

If that were the case, we wouldn't be able to detect the cosmic microwave background (all photons were emitted when the universe cooled enough for atoms to form). It would mean that the plasma era was longer than ~300,000 years and would cause more trouble for the Lambda-CDM model.

No. All LLMs have a setting you can tune (if using the API or running locally) or have it set by the provider and not tunable if using the chat interface. This temperature settings changes how the next token is chosen and since both input and output tokens influence the next one the answers might or might not start with the same words but will diverge, so the facts (there is no such thing as facts in LLMs, there are related abstract concepts joining tokens in the latent space) included in the answer will also differ.

Windows XP didn't have a compositing window manager and instead sent WM_PAINT messages to apps to repaint themselves when they became visible again. That's ~30MB per window extra RAM/VRAM required for Vista and newer. So let's say ~512MB extra with 15 windows. Next, with Windows 10, Microsoft stopped optimizing their disk access for HDDs and that made it almost unusable on HDD with the same amount of RAM as 7. Next, Microsoft stopped developing their Trident/Edge web browser engine and switched to Chromium that uses a lot of RAM to store all the JIT-ted versions of JS loaded. Then came all the Electron apps that don't share the same chromium engine but instead each has it's own copy in the RAM. Then came React/React Native and it's duplicate copy of DOM that makes it faster at the expense of RAM.

What is clean coal? It can be cleaner than existing coal based powerplants, but I don't think it can be anywhere as clean as renewables, hydro and nuclear.

Skype is for a long time now no longer P2P, all calls are relayed through Microsoft.

Corruption is unrelated to capitalism, communism or any other social system. In capitalism, corruption is means to become richer. In communism, it's means to become "more equal" than others.The end result is the same.

It's just that cabling needed for street lights (200W) and charging EVs (even slowest trickle charge, 120V 12A = 1.5kW) is completely different.

They only need to add "do not hallucinate!" to the prompt.

Thank you, this is some real data. 273g/mi vs 296g/mi, that's really close. Also, it is now 16%, but it looks like most of it got replaced not by renewables but by natural gas.

I'm actually curious, can someone explain to me how EVs reduce CO2 emissions? Worldwide, most of the electricity is produced using coal, and I don't see that changing for at least 20 years.