Feelings do not belong to software development anyway. Not in the past nor now.

The best (lowest) error rate was 17% in a test like:
Answer [this question] based on the data in [this pasted text].

Not sure whether it can be called awful but it is rather bad. I would not use it.
I did not save the reference. I only wanted to put some number so that people have an idea how bad it is now.

Not logging in and deleting cookies is the reason it works fine (without ads).

MS Intellimouse Optical is the only good MS hardware. I still use it. Well, they had also a nice force feedback joystick. Their keyboards are in the the "meh" category: no split, no thumb clusters.

Microsoft Natural Keyboard is not that natural neither ergonomic. Get a proper ergonomic keyboard. Something where left/right sides are split and adjustable and something which has thumb clusters.

Beside the already mentioned.
Sometimes you can order your resources and lock in that order. That rules out deadlocks.
You can describe with a petri net and reason or simulate over it.
You can also undo deadlocks with watch-dogs or transactions.
As for as use-after-free: well, use a programming language with a linear type system or with a garbage collector (GC). Linear type system is limiting but it prevents it. GC will transform it into increased memory usage which is easier to diagnose.

I agree here. The proofs are done against a specification. Some parts of a specification are easy: e.g. no null pointer de-reference. But these simple things can be often solved by using a more strongly and statically typed language. One can encode a lot of properties in types (of a dependently typed language if needed). The advertised AI system likely allows to prove even some non-trivial but still rather simple properties automatically.

The more complicated parts of a specification are harder than the implementations of the specification in a programming language. An important job of a programmer is to fill in the missing parts of an incomplete specification in a way which is the most likely to be actually required.

I did some proofs in COQ mostly for fun a long time ago. I enjoyed it a lot! But no expert in this :)

It is good enough.

The problem is that good comments describe why the code is written as it is. Good comments do not describe what the code does. That would be a duplication. What the code does is easier to see from the code directly.
Are good comments enough for this sort of technology?

Maybe what he meant is not that they [researchers making ChatPT] do not know how the model works and its limitations in general but they do not know a specific algorithm the neural network may be using.
Let's say they teach ChatGPT to properly multiply up to a given number size (without memorising it all). Would they know the internal algorithm the neural network is using to multiply? It is embedded somewhere in the weights but it may not be easy to read it out in a nice form readable by humans. This is in contrast to a nice multiplication algorithm we learn in schools.

OK, I did not know that influence on the chemical properties is big enough to be useful. The point of smell is that the sensors measure also vibration frequency of the molecules. This clearly depends on nuclear weight. At least, that is what the documentary claimed (but no claim of quantum entanglement role of course):
https://www.youtube.com/watch?...

Do you believe in quantum mysticism in our brains? Because quantum mind theory is not validated yet. Why current (classical) neuron models cannot do it all?

We know that our smell sensors are not purely chemical because they detect different isotopes differently but I never heard about anything like that as for as our brain is concerned.

I think it was a traditional fight between CEO and the board. In this instance of the fight, the board was concerned about safety; the CEO was concerned about money. The CEO did win.

It may be different from what MacMan exactly wrote but the consequence of closing nuclear plants ahead of their maximum possible life did result in lignite/coal emissions which would not need to be there otherwise. From this point of view, he was right. And his claim is still fully true in 2022.

Germany made a mistake. They started closing nuclear plants before coal plants. There is no way to spin this positively by pointing to the renewables in Germany. Politics sucks.

... should have written from 2019 onward ...