I predict his prediction will come true right around the same time he arrives on Mars. I'm really looking forward to one of those.

This shows why the AI spending boom isn't sustainable. One one hand, the only barrier to entry is willingness to spend insane amounts of money. Which certainly is a barrier, but there are enough large companies willing to do it that the market has become very competitive.

And on the other hand, there's no ceiling on how much you have to spend. The only requirement is, "More than your competitors." It's an arms race. Every company has to keep spending more and more to stay competitive. The spending grows without limit and the benefit they get from the spending doesn't.

The best case for them is that it becomes a war of attrition. Companies drop out as they decide they can't afford the spending. Eventually it narrows to just a few companies, and they agree to stop spending so much and share the market.

The worst case for them is that it all gets disrupted by other companies (e.g. DeepSeek) that figure out how to compete without spending so much, and their massive investment becomes worthless.

If you consider everyone in the business world, I think the rate is much lower than that. But if you consider only people in senior management positions, 3% sounds about right to me. That's based on people I personally worked with who later did or were found to have done illegal things they could have been sued or prosecuted for.

But none of them actually was sued or prosecuted. There would have been nothing to gain from it. Once people figured out what they were doing, the response was to show them the door and try never to see them again.

The same is likely true of the Forbes list. If 3% get charged with crimes, the number committing crimes and not getting charged is probably much higher.

That's why they're focusing on middle mile, not last mile. Last mile is hard. Your truck needs to go anywhere, and you don't know if there will be someone there to unload it at the end. Middle mile between warehouses and stores is a lot easier. It only needs to drive fixed routes that you assign in advance, and you know there will be someone to unload it at the end.

Why not focus on long-haul highway driving? Because other companies are focusing on that. This is a different market that requires a different product.

In a post that it will soon be illegal for anyone under 15 to read.

Leaders need to practice what they preach. If you're going to cut teens off from social media, don't put important information on social media where they can't access it.

The post you're replying to said Christian Nationalist, not Christian. They're not the same thing. See the Wikipedia article which begins, "Christian nationalism asserts that the United States is a country founded by and for Christians."

That's not how venture funds work. The company managing the fund (Mozilla in this case) doesn't invest their own money. They find investors who want to participate, and get a share in all the companies they invest in as payment for organizing everything.

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!