Presidential immunity and panic

I'm not joining in our collective, competitive national panic attack. I do not think our republic is dead or doomed. It will right itself. Considerations:

• The president can still be impeached. Impeachment is the normal way of dealing with crimes committed by presidents.
• The immunity granted to Trump has also been granted to Biden. Some people on the right wing are panicking at this thought.
• The immunity is only for official acts. It's pretty clear that the crimes Trump is accused of were not official acts. (Do I recall that part of his defense in the Stormy Daniels case is that he was not acting officially? Can't have it both ways.) I think we are going to see some court cases that will quickly show how narrow the immunity really is.
• Hearings to determine whether actions were official will in effect be preliminary trials that put presidential misdeeds, and the evidence for them, before the public sooner than the actual trial.
• Only the president is immune. For everybody else, obeying an unlawful order is still illegal. And while such a person could perhaps be pardoned by a corrupt president, it would be a mark of shame, a career-ending move.

So... We've had a glitch; it will be corrected; and in the meantime, I have only one thing to say:

Trump supporters, this is your circus and these are your monkeys. Think carefully about how you take it from here.

Permanent link to this entry

Artificial intelligence is not omniscient, omnipotent, or necessarily superhuman

It is a serious mistake to think that if computers model or emulate human thinking, they will necessarily do it with superhuman speed and accuracy, or that "artificial intelligence" can do the impossible.

Computers are certainly superhuman at things like arithmetic and routing and delivering messages. That's because they're machines. They don't do these things the way humans do; they use their special-purpose mechanisms.

It is quite possible for a computer to model a human capability with sub-human performance. Think of speech recognition and face recognition. What if "artificial general intelligence" arrives but has an IQ of 70?

Sometimes AI techniques have inherent limits. ChatGPT is trained to model the way words are used in context. People use it as a repository of knowledge — and it gives inaccurate output because it paraphrases things in ways that are grammatical and natural but not truth-preserving. This would be the case even if the entire training corpus were carefully screened for accuracy. A bigger, faster computer would not change this.

And AI cannot do the impossible. Someone asked me when AI was going to give us perfect forecasting of the weather and the stock market. Never — because those things can't be done! The information needed is simply not available in full. And AI does not somehow omnisciently get information that there is no way to collect.

Permanent link to this entry