It's not a moving target at all. The term was coined in 1955, and its accepted definition hasn't changed the slightest bit in the 70 years since. It comes from the proposal for the Dartmouth Summer Research Project on Artificial Intelligence. Here is how they defined it.

That is still exactly what practitioners in the field take it to mean. If you come across other definitions, that's people making up their own definition because they don't know how the field has defined it for 70 years.

Two important points about this definition. First, artificial intelligence is defined as intelligent behavior. If a machine behaves intelligently than it is intelligent. How that behavior is implemented doesn't matter. Second, if a machine can do anything a human can do then by definition it's as intelligent as the human.

These ideas were strongly influenced by Turing's "imitation game" (now known as the Turing test) that he published five years earlier.

I challenge anyone to come up with a definition of intelligence that 1) is clearly satisfied by humans, 2) is clearly not satisfied by current AI, and 3) isn't totally contrived (like defining intelligence as "the ability of humans to...").

Here are some common definitions of intelligence.

"The ability to take in information and make decisions based on it." Computers do that.

"The ability to solve problems." Computers do that.

"The ability to predict the consequences of your actions." Computers do that.

"The ability to understand complex subjects." Humans often speak or reason about things without really understanding them. For examples, see the comments thread on any slashdot story. Besides, "understand" is another word that doesn't have a clear definition.

If you believe AI isn't intelligent then how do you define intelligence? Are you absolutely certain AI doesn't satisfy your definition? Are you absolutely certain that humans do? And if you answered yes to both, is it a contrived definition that doesn't match how most people use the word?

I hope you meant that as a joke! If not, it's wrong in so many different ways.

The difference between data and anecdotes is that data is sampled from a probability distribution. By looking at the samples, you can learn about the distribution they're drawn from. That lets you make predictions about future observations drawn from the same distribution.

Anecdotes are chosen individually. They don't reflect any probability distribution. They don't let you predict future observations.

It goes deeper than that. For many years, the air force literally told officers that UFOs were real, that there was a secret program to reverse engineer them, and that they would be arrested if they ever said a word to anyone. It was basically a hazing ritual. This article has a great discussion of it. Here's a bit of it.

Not quite as bad as that. Average power plant efficiency is now about 39%. Cogeneration can further improve it by capturing the heat and using it for useful things.

ICE cars are even worse though, sometimes as low as 20%.

This is also why the transition to renewables is so important. If your generator doesn't use fuel and doesn't produce CO2, the question of how much fuel it wasted or how much CO2 it produced is moot.

I just searched for "cost of energy sources", and it took me less than one minute to find the data. Here you go. There's a graph right at the top helpfully showing you how they compare, and how the costs have changed with time.

I bet you could have found it just as quickly if you'd bothered to look. See my signature quote.

Electricity is the wrong number to look at. Total energy is what matters.

If you replace an ICE car with an EV, you'll use less energy but more electricity. Same if you replace a gas furnace with a heat pump, or a gas stove with an electric one. Looking only at electricity makes all these things look bad, when in fact they're good.

In an accident you don't need the windshield to resist breaking. You need it to fragment into tiny bits that can't impale you. Auto glass is well designed for the purpose.

Let's stick to technologies that might actually be possible, not crazy fantasies!

There are different levels of rare. Lithium isn't platinum, but it's not sodium either. I keep a tub of salt (50% sodium) in my kitchen cupboard. It's dirt cheap. Lithium can't compete with it on price.

Sodium can't match lithium on mass, but like the article says, the bigger use is for grid storage. Weight doesn't matter much if you aren't carrying the batteries around with you. Using sodium for grid storage makes it cheaper, and frees up lithium for cars.

To realize just how insane this is, consider that we had weak La Nina conditions for much of 2025. That means it should have been a slightly cooler than average year. And it was the second hottest year ever recorded.

That tells you just how quickly the climate is changing. 2023 had a strong El Nino. A cooler than average year today is as warm as a hotter than average year two years ago, and much hotter than an average year was three years ago.

Not even close. The US is number 16, with per-capita emissions less that 1/4 of the worst country. All the worse countries are smaller, which is why the US has higher total emissions. They aren't tiny though. Russia, Canada, Australia, and Saudi Arabia all have higher per-capita emissions.

Does someone not understand what solar means?

I think if a company buys a 100 horsepower engine, they should report that they added 100 horses to their stable. It's basically the same thing, isn't it?

I guess maybe renaming the company for the metaverse wasn't such a great idea after all.

Your information is out of date. Read this if you want current information. Most EVs now have similar range to typical ICE cars, and the effect of cold on range is similar for both EVs and ICEs, usually about 15%.