I know due process only applies in criminal cases, but this seems egregious:

It seems like the fight goes back and forth about "just cameras" vs "cameras + sensors", when all I can make of it is that the REAL answer lies in how good the hardware and software is processing the incoming data?

Tesla, for example, scrapped the cameras + lidar idea, largely out of frustration that both inputs often led to the computer interpreting the scene in conflicting ways. "Camera visual seems to indicate that's just a light pole on the side of the road but lidar image shows it may be a person starting to walk into the road." (Maybe that's a poor specific example, but I heard talks about it saying things along these lines were the struggle.)

If Waymo is arguing humans can only do so much with our eyes, so "superhuman levels of driving" require more? Again, it seems to me like more CPU power and better software is that "more"? Computers can do math FAR quicker than my brain can... and that's been true for decades. Superhuman levels of driving safely seem like they could happen with systems just using cameras OR systems collecting data from multiple sources?

You're right, but the problem is going to be location in many cases. Would a former data center make a great public library? Technically, it might. But when it was built out in the middle of some farm land, where they originally got a good deal on the land by buying some farmers out who were willing to sell? Now, it might not make so much sense.

Or as an example closer to home for me? Our own city is fielding requests to construct data centers right now. There are rumors about our now closed-down steel plant being a potential site. Great for the guy who owns the property and wants to unload it. But a data center there would mean tearing out and removing an existing environment that was good for manufacturing. If you said ok -- no more steel made here but we could do cement, or process glass? That would be an equivalent replacement that would bring back the lost jobs. A data center really won't ... and if it closes? I'm not sure people would want it as office space in that part of town.

With their COVID vaccine, the first time was the highest dose they knew would have only acceptable side effects, and they reduced it for the next iteration when they saw how effective it was. (If you remember the myocarditis rare side effect, that was only seen with the first version.) I would guess that the next step is backing off on the dosage to reduce the side effects now that they have results on effectiveness, which will also let them make a better case for younger people at less risk from the flu.

Lidar doesn't guarantee things, but since you brought up fog, guess what *can not* sense well in fog? Cameras. LIDAR and RADAR do better in fog.

Yes the problem with self driving goes beyond having the right sensors. That doesn't mean you shouldn't have the right sensors.

He's not lying, he's simply stating that you can give automation super human senses, and it would just be a complete idiot that would be so against those senses even in principle that hehwould remove common radar sensors from his car, just to prove a point.

He's not talking about worst human case or maybe even typical human. He's saying that confining the solution to cameras makes it only as safe as the safest human, which is fair.

Radar is pretty cheap for cars. LIDAR has similarly gotten pretty cheap. Being anti-LIDAR had at least some justification when LIDAR could only be had for thousands of dollars and saying we can't afford it. Being anti-RADAR was stupid from the moment that sentiment was expressed.

This is a stupid waste of time that could not have *possibly* earned them that much, a single ad in a very narrow context to a fairly small demographic. They risk their reputation on what has to be a relative pittance. Same on when they toyed with 'heated seats as a service'. A demographic that was probably happy to pay, but the annoyance was enough..

Similarly, GM has decided to move toward no support for phone interoperability. They are de-featuring cars that might be over $90,000 in the hopes that *maybe* people will accede to $10/month for the features their phone gives them already. Risking over $10k profit on a unit to have a somewhat higher chance at an additional $10/month.

I demand at *least* "Ultraintelligence". How can I settle for "Superintelligence"?

Reality is? The places that allow new data centers where they interfere with the existing population are just going to wind up with regrets.

They remind me a bit of the battles over putting in new landfills/dumps. Everyone knows we need them, but it's always "not in my back yard"!

The thing is? The need for more data centers is questionable. If all this AI growth turns the corner and companies decide much of it was just overblown hype? Then those data centers will quickly wind back down again.

Even in "best case" scenarios, they don't employ many people once they're constructed, and they use large amounts of resources -- which tend to be sold at discounted rates as part of the deals struck when they move into an area. Everyone else gets stuck covering any losses by way of rate increases.

A landfill may smell bad, but it's not using up the lion's share of available electric power in my town or using a big portion of available clean water.

I don't understand how the entire concept isn't just outlawed and thrown out? The core reason we have a stock market is to allow business growth/prosperity. Company owners trade off their original ownership of a successful business for a chance to get investor income flowing in to make the business more successful than they could make it on their own.

Prediction markets are nothing more than gambling. There's no "investing" in a business happening at all with this.

To be fair, cellular data is really poor in quite a few places in America, when using ANY of the "second tier" reseller carriers. (Won't speak for other countries.)

EG. We have Spectrum Mobile right now (mostly due to a bundle deal with our broadband at home), and while voice calls work pretty much as expected, anywhere I go? The data is terrible! They say they're using Verizon's towers but it's obvious Verizon is throttling the heck out of them and/or only allowing them on SOME of the towers in a given area.

When I had T-Mobile for a few years previously? It had bad holes in coverage areas, but it worked quite well where it was built out (usually your bigger cities).

I don't care how much "cost savings" they claim it brings to eliminate retail stores. As one of only a few "major carriers", it's pathetic to offer only online interaction with customers. And honestly? I'm not even sure I would have a big issue with the FCC pulling some of the dedicated frequency spectrum they were allowed to purchase. I'm pretty sure there was an assumption made that as one of the "big 3" cellular providers in the USA, T-Mobile would ensure any potential customer had the ability to buy a phone with them and get the help needed to use it. I imagine there are quite a few elderly or handicapped people around who don't find it very functional to do all of that via an app.

Or give them a budget and don't micromanage how they work but instead evaluate their results.

Problem is that management would actually have to understand the work and the business, and I suppose that is just too hard.

I don't do a lot of traveling, but I remember when I last worked at a job that involved some travel? The ongoing "wisdom" was a realization that the cheaper hotels had the most usable/reliable wifi. The expensive ones tended to use Internet as a service to upsell, with annoying portals trying to bill you a daily rate and generally interfering with your applications. (They might, for example, lock down just about all ports except 80 and 443, meaning you couldn't get software updates to complete or connect a VPN, or ....)

The basic chains, by contrast, all seemed to used "canned" wifi setups sold to them by third party vendors, and as long as you got a decent signal in your room, you were good.

In any case, I never trusted a hotel wifi to be super-secure. Heck, I remember on a Vegas trip one time, they had it set up so wifi was capped at a slow rate unless you paid a premium. I was curious about a wired jack I found behind the bed and plugged into it. Turns out, it gave me a full gigabit network connection and bypassed their sign-in portal and everything. They seemed to just not realize or expect guests would still carry ethernet cables with them and try to use a wired connection.

Even if open, it just wasn't going to happen.

Problem was that it was awkwardly slower than DRAM and more expensive than NAND. Was it cheaper per GB than DRAM? Sure. Was it faster than NAND? Sure. Problem is there wasn't really space in the market for a tier between NAND and DRAM. Embarrassingly obvious as Intel ran around doing things like trying to force a whole different paradigm for application developers to do things awkwardly special to try to take advantage of this tier.

Now we have a scenario where one would think Optane could shine, an industry obsessed with feeding AI models as fast as possible. Except the answer is simply "more NAND in parallel" rather than fundamentally distinct technology.