AI has no moat.

Except for the people actually studying that kind of thing, of course.

A very important work in that space is Gilbert Ryle's Theory Of Mind, 1950. It introduces the concept of "category error" to explain how people like Decartes confuse a metaphor for what it stands for. ("To the left is the forest, and to the right are the trees of the forest." That kind of thing.) It also lists ways in which the mind is not a continuous thing, and in particular how consciousness is only present occasionally. He thoroughly disproves the Cartesian body-mind dualism.

Also interesting is Daniel Dennett's Consciousness Explained, 1991. Despite the title, it is more of a tacit proposal than an explanation, but it draws from a vast corpus of psychological experiments about perception and inner monologue, and dispells common and persistent myths about the mind, consciousness, and perception; in particular what he calls the "Cartesian theater of the mind". Like Ryle, Dennett emphasizes how the mind is not a physical locus, but more of an ephemeral phenomenon, and not a central observer or actor either. He introduces "heterophenomenology" and the "many-drafts-model" to explain how perception and innner monologue can change from post-hoc stimuli without the conscious observer being conscious of this. He presents some fundmental criteria for consciousness. (Although he misunderstands computer programming.)

The question how consciousness can exist in the physical universe is misguided, at best. Without a physical universe, consciousness could not exist at all. This is something both Ryle and Dennet address as they dismantle Cartesian dualism. (The idea that consciousness exists independently of the body, as Descartes suggested, is a lie that is required only by religion: An afterlife or reincarnation requires that you are still alive when you are dead. Meanwhile, in reality, the illusion of continuity of consciousness is only an illusion, created by the conscious mind not being there to observe when it is not there.)

A lot of work has also happened in neuroscience, but I'm not up to date to the latest research.

Google did it since 1997.

Sorry. When you signed up for that discount card a few years ago, you agreed to individual binding arbitration only.

> It can handle Multimedia as well as OSX

Found the Slackware user. The 32-bit Slackware user. The 32-bit Slackware 3.x user.

Have you used GNU/Linux since, say, 9/11?

Apple Hardware before 2012 cannot run operating systems other than those branded OS X or Mac OS X. While, technically, anyone who bought hardware before 2016 or so could be sticking with the OS that came with their machine, anyone with such a machine will get nagged and will not be getting updates to Safari.

The more likely explanation is this is bot traffic using older UA strings and SC's figures are BS.

OS X has been macOS for a while, and Apple, alas, makes its hardware obsolete after about 5 years (say one thing for Microsoft, they usually - recent TPM/Win 11 stuff aside - allow much older hardware run its latest OSes), so I'm wondering what's going on there. Are there really a lot of users of Macs from 5 years before whenever the rebranding started in 2016? That means a lot of people using Macs from before 2011. A huge number. Which genuinely surprises me. I have a Mac mini from 2014 and it crawls to the point of being borderline useless

So... maybe "OS X" is referring to something else? Maybe this is bot traffic using older user agent strings? I've never trusted StatCounter, I used to manage a set of websites that had millions of users across a relatively broad set of demographics, and what we were seeing via Google Analytics didn't remotely match what they were claiming. And by not remotely, I mean not even in the same ballpark.

The idea that they're not even checking whether the traffic is bot generated wouldn't surprise me.

they could put a photodetector next to the LED

Which you could replace with a resistor connected to where the LED used to be.

use the LED itself as a light detector.

Ditto.

It's difficult to defeat someone who's moderately competent and has access to the hardware.

It certainly is good that any of the properties of an LED that you can measure cheaply and reliably enough to get away with using in consumer electronics are 100% distinct from those of other components or precisely the same LED covered with opaque epoxy.

Just detecting that there's now an open circuit where a diode should be would be fairly trivial and cover the cruder drilling cases; but this will be cosmetic at best against any moderately motivated tampering.

You might be too young.. oh who a hell am I kidding this is slashdot.

Anyway in electronics there are different grades of quality. TigerDirect made a business selling the B and c grade stuff. Because of that they could undercut just about anyone but they often sold junk.

Think of it like how Intel bins their CPUs only it's stuff like the quality of the electrical wiring in the final product.

So the practice I think has more or less gone away for major brands. They have much better quality control so they don't do the binning thing anymore. But the factories will absolutely do binning and resell them to different brands. So there can be a big difference between the Kodak electric scooter and the one from some weird Chinese seller. Although honestly kodak's brand is so shot these days that I doubt there's much difference between those two particular companies

Apparently broadcom does have fabs and they make chips and apple uses them for that. I don't really think of broadcom like that but yeah I guess they do. So it's Apple designed chips being manufactured by broadcom.

As a matter of national security yeah it's good to have chip production here but it's kind of a yawn from a jobs perspective. None of these chip foundries produce very many jobs. It beats the hell out of an AI data center I guess.

That sounds like the journalist getting confused. The "supertorquer" seems to be a magnetorquer that uses superconducting coils. You can get propellant-free orientation control that way, but not propulsion. It's right in the name: "torquer."

You can get propulsion, sort of, from a magnetic device like an electrodynamic tether, but it's a different thing, and would be very, very hard to make superconducting. If you wanted to go to Mars using magnets you'd use them to build something like a magnetohydrodynamic drive, which aren't propellantless but can be enormously efficient and are likely to be one of the many things that benefit from the new high temperature superconductor ribbon.

The whole thing sounds pretty arbitrary. I've got an electric screwdriver with some name like "Goofun" on it that was clearly designed by engineers who wanted to make the best electric screwdriver they could. I've also got one from "Black and Decker" that is of much lower quality.

WTF is "Black and Decker" anyway? Or "Google?" "Microsoft" sounds like a sex joke, and last time I had "Coca-Cola" it didn't contain any of either.

I like to link the venn diagram that has the large circle AI with AI (but non-ml) application

Yes, I think this is what leads to a lot of the confusion. "Artificial intelligence" really refers to the application. It's properly the study of problems that humans tend to be good at and classical machines are not. Arithmetic is not an AI problem. Image and video processing, pattern recognition, planning tasks, game playing, natural language, stuff like that is.

Machine learning refers to a class of techniques that are (most often) applied to AI problems. The first efforts that way were not ML though, the idea was to convert the problem into arithmetic, normally via a logic system. Prolog is a language that was designed for classic AI but doesn't really do any machine learning.*

Making a Venn diagram with AI and ML is like making a Venn diagram with "Cleaning the Kitchen" and "Brands of Paper Towel." The members of the sets are related, but they're not the same type of object. I suspect what you're actually imagining is a Venn diagram "Techniques used to solve AI problems" with subsets for "Classical" and "Machine Learning", the latter of which has (overlapping) subsets like "Neural Networks", "Deep Learning", "Statistics", etc.

* there are some modern machine learning systems that provide their output as a prolog program, but you don't do the learning in Prolog.