fl7305

> So that wall was located at the beginning of the runway if the runway was used correctly.

Most runways are intended to be used in both directions depending on the wind. This one doesn't seem to be an exception?

What should ATC have done differently?

Historical note: Stalin and Hitler agreed to start WW2 by invading Poland in September 1939.

So Stalin may have played a part in ending WW2, but don't forget his part in starting it.

Not sure what you mean by "LLM can't do Socratus' method"?

But: Plenty of people struggle with playing along with Socrates' method. Can they not reason at all?

Neural networks can be evolved to do all sorts of algorithms. For example, controlling an inverted pendulum so that it stays balanced.

> In a CPU, we take great care with data representation and protocol definition, then we do formal verification on the algorithms, and we can be pretty sure that the output are correct.

Sure, intelligent design makes for a better design in many ways.

That doesn't mean that an evolved design doesn't work at all, right?

Can a blind man sculpt?

What if he makes mistakes that a seeing person would never make?

Does that mean that the blind man is not capable of sculpting at all?

> It's claimed that this model "understands" chess, and can "reason", and do "actual logic" (here in the comments).

You can divide reasoning into three levels:

1) Can't reason - just regurgitates from memory

2) Can reason, but makes mistakes

3) Always reasons perfectly, never makes mistakes

If an LLM makes mistakes, you've proven that it doesn't reason perfectly.

You haven't proven that it can't reason.

> The internal model of a LLM is statistical text. Which is linear and fixed. Not great other than generating text similar to what was ingested.

The internal model of a CPU is linear and fixed. Yet, a CPU can still generate an output which is very different from the input. It is not a simple lookup table, instead it executes complex algorithms.

An LLM has large amounts of input processing power. It has a large internal state. It executes "cycle by cycle", processing the inputs and internal state to generate output data and a new internal state.

So why shouldn't LLMs be capable of executing complex algorithms?

Has anyone done good power measurements of similar boards?

How many watts does it draw in idle, surfing the web, and watching a YouTube video?

> We don't have any more ability to test a CO detector than you!

I'm sure you're technically correct.

But if you want to, then it's easy to test CO detectors. Industries do it all the time. You just buy spray cans with known CO concentrations ("bump gas") and check that they go off when sprayed.

> Please don't call 911 unless you have an actual emergency.

Sure. But if the city decides to prioritize fire safety, then CO detector testing would be a good preventative measure that can be offered to the local residents.

Good CO safety is more than just the detector working. It's also about placement in the room, how many in each apartment/building, etc.

On top of that, the more the fire department interacts with the local residents, the more they can educate them about all kinds of fire safety matters (do you have a fire blanket in easy reach in the kitchen when cooking with oils?).

It also provides a relaxed way of fixing fire hazards (please don't stack unused furniture in front of the fire exit).

In Germany, very often. VDE is like ANSI or IEEE there.

If you ever work on anything electrical in Germany, you see the acronym VDE everywhere on standards etc.

It's an interesting idea, but I think you'll be surprised at how fast the sun moves. You can try it yourself. Put up a small mirror that reflects a sun spot on a surface. Clock how fast it moves a centimeter. Multiply that distance with how much further away a central tower would be.

For concentrated solar polar, you need almost constant movement of the mirrors to keep tracking the sun well enough.

So you need independent azimuth+elevation for each mirror.

It's not great for a production site. But for engineering tests, you can do a lot of stuff with no sun, and the occasional clear skies are used for sun testing.

You normally keep all mirrors pointed at the real target. Then take one at a time and point it at the black board off to the side.

You can then use the camera image of the position of the sun spot on the black board to figure out your new calibration settings for this particular mirror.

> fly a scanning pattern around the focus point and look for mirrors that blind the drone instead of frying the tower.

Or just put up a big black board with reference points on it, and point a camera at that?

In my view, it sounds like they're trying to tackle way too much at the same time.

Just the core system that converts input chemicals and heat to a fuel is a major undertaking. Focus on that.

Adding their own mirror system might be doable, but only if they use a well known and simple solution.

Trying to add an "AI drone mirror autocalibration" is a major undertaking, enough to keep a medium sized company busy for a number of years.

Likewise with the "solar energy storage system". Just run when the sun is shining, and produce as much fuel you can from that.

If you have more incoming power from the sun than what you can produce fuel from, concentrate on solving that instead. Or just build something that is economical even if the mirror system is way oversized in order to always keep the fuel conversion busy.

1,500 C sounds like it's way past what any normal metals can handle?

But in this case, maybe it is enough to have just a small part of the central reaction vessel handle 1,500 C? Maybe using ceramics?

Or can you 3D print an Inconel chamber with builtin cooling channels that makes it possible for the Inconel surface to handle the incoming heat?

>> And finally, a thermal store to release energy when the sun goes down to allow the solar-powered facility to operate around the clock.

> This actually IS new and interesting in this application

Yes, it's a nice thing to have. But it is a major research and engineering project just in itself. It is by no means a solved problem where you can buy a working solution from someone.

> This bit smells like trying to shoehorn in an application of "AI" where it's not really needed - what's the actual improvement using "drones and AI" over just pre-calculating a tracking curve based on latitude + time of day/year? Or just putting down twice as many mirrors and not bothering to make them track?

Open loop (just pre-calculate) is pretty inaccurate in a system like this. Especially over time as the mirror positioning degrades with accumulating inaccuracies.

So there might be a business case for their AI drone calibration system. But just that part in itself is also a major undertaking that requires a significant investmet over a long time to make it a production system.

"Just add more mirrors" sounds like a better business proposition, coupled with a much simpler autocalibration. Perhaps an off-center bullseye target where each mirror periodically aims itself, and a central vision system figures out the azimuth/elevation offsets?

> This is hyperbole

I agree in that I'd be very surprised if they ever get to the point of mass production.

But I agree with the point that generating synthetic fuels from the sun in a 100% renewable way would be very significant.

Sure, in 30-50 years electrical transportation might meet 99% of our needs. But we're not there yet.

> Naively I would assume we know where the sun is and how to calculate the angles. What are drones and AI doing?

Yes, you know exactly where the sun is in a global coordinate system.

The problem is that your mirrors are not fixed in that global coordinate system.

You need to continuously re-calibrate the local coordinate system of the mirrors against the global system.

Really? Looking at the asking prices on eBay, a VT100 terminal seems to go for many hundreds of dollars even in poor condition?

You wouldn't happen to have SUN 3 workstations?

Like the other poster wrote, it is time dependent.

The stuff that went into dumpsters in the 1990s can sell for thousands today on eBay. But in 30 years, they might be dumpster bound again.

I just double checked to make sure, and an Atari Falcon early 1990s 68030 CPU computer sells for around $3000 today.

$10k sounds high to me if we're talking about home electronics, furniture, etc?

Do people have a lot of expensive jewelry that's included, or does it include vehicles?

Well, I see classic 1970s-80s computers that I'm interested in selling for hundreds or even thousands of dollars on eBay etc.

Gen-Xers who grew up with them are getting nostalgic, and many of them have a lot of disposable income and plenty of space to house them.

But it is a lot of work to get that kind of money for each individual machine. And I suspect it will be a passing fad. In some decades when the Gen-Xers start to log out, the younger generations will not sustain the same kind of market.

The Swedish company Pricer has done this since 1991.

> Ag’s should not be requesting that this is fixed, but requiring it in no uncertain terms, and giving out massive penalties for every single time it’s allowed to happen.

Or just write up criminal charges and arresting the CEO of Meta the next time he sets foot in their state.

A weekend in a county jail might realign his priorities?

Maybe mount the display on a small adapter PCB that has a CPU for pixel conversion, and a standard flat cable down to the main board?

Then the main board can stay the same even as the displays are swapped out together with new adapter PCBs?

I know, just spitballing :)

Use a RP2040 or similar to handle the conversion from the Spectrum pixel output to what the display needs?

And perhaps a small adapter board for the display FPC connector?

On that note, 17th century Sweden can report that moose make for very poor cavalry steeds.