Yes and no.

I agree that it's likely that there's no specific line, at least not a sharp one, but there is a qualitative difference between machine learning as we know it now, and human learning. Human learning, at least the best human learning, is about the creation of knowledge, not the acquisition of facts, nor even the identification of key facts from a larger mass (which is what machine learning as done today is about).

A key difference is explanation. A human playing Breakout not only comes up with the strategy of tunneling and exploiting it to knock out lots of blocks from the top, but can explain why that strategy works. This sort of meta-learning, knowledge plus understanding of knowledge, is something that none of our algorithms can yet implement.

This point I don't know about. What you're saying is fundamentally religious, ascribing to people some element that makes us more than the mechanical and chemical interactions that comprise us. That may be the case, or it may not. Even for those who believe in God, it has proven tricky to assume supernatural explanations, because physics often finds natural explanations for them.

Ah, I see what you're saying. I agree. The original poster's claim that Corney was attacking the Constitutional right of privacy. He wasn't, since that right is removed (in specific ways) by a warrant, and he's saying his concern is that this technology makes warrants ineffective.

On the other hand, nothing in the Constitution requires that we go out of our way to enable warrants to be effective. Given that encrypting devices is a very reasonable precaution to take to protect our data in the case of loss or theft, there's no reason we should avoid it just to help out the police.

$ cat > test.c
printf("Hello, wold!");
$ gcc test.c
test.c:1:8: error: expected declaration specifiers or ‘...’ before string constant

Given that it doesn't even compile, I'd say it's buggy as hell.

Or rather than eliminating it, secure it. A couple of proposals:

http://www.certificate-transparency.org/ provides a distributed mechanism for near real-time monitoring of certificates in use, to very quickly identify and block certificates that weren't issued legitimately.

http://convergence.io/ makes the observation that MITM attacks result in some subset of the Internet seeing a different certificate from a given server. A distributed system of crosschecks identifies sessions which are being attacked.

Note that in spite of the fact that Convergence is billed as a replacement for the CA system, there's no reason that we can't use both.

Neither was the AC to whom you replied. Gameplay testing can and should be automated.

Oh, one more point: I should mention that it seems pretty likely that once we understand how human learning works, we will actually not program machines to learn the same way; we'll program them to learn a better way.

I think you're probably right... but we can't prove it since we don't yet know how human learning works. Once we do know how human learning works, we will be able to program machines to learn the same way.

And all of their personal and business correspondence, and access to their bank accounts, brokerage accounts, password managers (though access to someone's e-mail is generally sufficient to get into everything else on-line), etc. Your phone can also tell someone where you go (navigation history) and if you have it turned on can even provide them with a detailed account of where you have been, every minute of every day.

The aggregate content of a smart phone is, for many people, everything about them worth stealing.

Predictive power is necessary, but not sufficient.

I don't agree that it's "the most fundamental criterion", because it's not a matter of weighing various criteria, some more important than others, and saying "yep, this theory is good enough, it's a keeper". Instead, all of the criteria must be met. An explanation must make verifiable predictions, those predictions must be borne out, the explanation must not be easily variable to accommodate any observations and must imply its own reach.

As to your example, QM doesn't say that systems only behave quantum mechanically below a certain scale. They behave that way at all scales... it's just that at larger scales the collective probability of observing the quantum "weirdness" is so low that it effectively never happens. At least in most cases: Young's experiment shows that QM behavior can be quite apparent at larger scales.

FWIW, I don't consider QM to be an ugly theory at all. In particular, I find the many-worlds interpretation of QM to be quite elegant and beautiful -- though still quite brain-twisting and very different from our common conception of the world. But then, frankly, classical atomic theory is also quite alien to our common conception of reality. For example, the notions that solid matter is nearly all empty space and that the fact that we can't walk through walls is actually due to counterposing fields not because the matter of my forehead and the wall really "collide". The notion that we live on a spherical body rotating around the sun is also quite counter-intuitive, frankly. It only seems clear and obvious because we're used to it. So degree of match with human perceptions of reality isn't really a useful yardstick.

Interestingly, I think beauty and elegance are useful yardsticks, though, because they tend to bear precisely on the "hard to vary" characteristic of good explanatory theories. The more complex a theory the less elegant and beautiful it is, and the more "knobs" there are to tweak to explain different observations. Current explanations of subatomic particle physics are deeply unsatisfactory... and not only are they ugly, with their explosion of seemingly random types of particles, we don't have explanations that tell us why we observe all of those particles and not others or (with some exceptions) what all of the other particles may be. This lack of reach and excess of variability indicates that we don't really have a scientific explanation of subatomic particle physics, even though we can make some predictions.

Assuming one can *find* said .22LR.

Not completely true. There is value in having developers who are from the countries where you sell your products because their understanding of the local culture and context can help them to design and build products better suited to the customer base. But assuming you address that issue (or just don't care about it), yeah, dev shops can be anywhere and everywhere.