This is an excellent treatise on both the value of a computer science education and the importance of being sure that things you say in public aren't verifiably false.
Verilog has many drawbacks, including no support for structured signals. On the other hand, the other big language, VHDL, is really difficult to use for "modular" projects. Is anyone here familiar with CHDL, a C++ hardware design language?
> Now, some people may argue that these algorithms are not examples of "intelligence". The obvious conclusion must be that hiring people, beating people at Go, and playing Super Mario must also not be tasks that require intelligence.
Or it could be more nuanced than that. Effectively we have shown that given enough time (and data, and clever learning algorithms) we can teach machines to answer questions that we tell them to answer. But that is only part of intelligence---which itself is a tricky concept to even define. An intelligent agent is able to pose questions that need to be answered. (Or at least, we can probably come to a consensus agreement on that statement.)
And yes, this is a minor nit about a careless statement that was made, not an analysis of the full article.
I think it's possible that "intelligence" is by definition always a chauvinistic property: it is whatever we possess that nobody else does. It is defined precisely by that difference between anything else, and "us" (for varyingly inclusive definitions of "us").
E.g. whales and giant squid are almost certainly intelligent in an absolute sense, relative to the average self-replicating organism (which is probably a bacteria or yeast at best, maybe an overgrown protein depending on who's doing the computation), but we frequently define "intelligence" in such a way so as to exclude them. I would argue that this says little about the quantity but a lot about those doing the defining.
"Intelligence" will always be out of reach for anyone other than humans because as soon as machines get close, we will simply move the bar a little. The nebulousness of the definition allows us to always maintain a je ne sais quoi that is all our own, and will never be approached by machine or beast.
A practical definition of intelligence is the ability to successfully abstract and predict the outcome of interactions with an organism's environment in ways that aid long-term survivability of the organism's species.
With this definition, a Go playing robot has extremely limited intelligence. It can't model anything except the Go game in front of it.
Any single-talent AI is going to have the same limitation.
Most mammals are somewhat intelligent. It's not clear if dolphins and whales do much abstraction, or even if they need to do much abstraction to survive. Some basic heuristics are enough to get by, and - so far as we know, which isn't much - life in the oceans is less varied and challenging than in all the many different habitats on land.
Even so, dolphins and other animals can learn and improvise, which suggests a basic ability to model and predict reality through internal mental representations.
Humans are modestly intelligent. We've learned how to do basic abstraction and social learning, but we're still stuck in a bottleneck of individual competition, and are only just beginning to consider the possibility that it could be useful to model and predict our future as a species.
I would guess - based on no evidence at all - that at some point an intelligent species breaks through the bottleneck and starts being able to model the interactions of the species as a whole with the environment as a whole. That likely means starting with planetary environments and terraforming, but for all anyone knows it may expand to something much broader - because a species will have to understand itself extremely well to survive successfully the bottleneck.
So a true multi-talent AI will have to model the physical world - i.e. understand and improvise its own science - and also do the same with the human world. Which means it would understand and improvise human politics, psychology, and culture.
That would be quite something - light years from the current machine learning classifier projects that are so popular.
It's worth remembering that the benchmark for that AI isn't any one human, but all of human culture as a whole.
Most humans can't play Go, never mind win at it. The average human is also pretty terrible at music, psychology, science, art, coding, and so on. Worse, the average human will take a very long time to learn any of the above, and will still be averagely mediocre, even after a lot of time and effort.
But an average human is still good at being human - i.e. understanding and generating, possibly improvising, correct social and conversational responses for a variety of set situations.
AI theory seems to get this ability confused with intelligence. In fact a single-talent social AI should be possible. But even if it appears averagely human, it won't necessarily be a good general self-directed multi-talent AI that can quickly learn to beat the best human in in any problem domain.
PVS-Studio (or more specifically Andrey) used to do a really nice program called "CppHints", where they would email a little tip each day (and later each week I think).
Anyway, I guess I just wanted to say, Andrey, I appreciated your work and enjoyed reading those tips! Thanks!
I had fun playing with this, but what I noticed is that the recommendations started well, but as I rated more movies, they became less and less accurate.
Is it possible that whatever technique you are using to model individual users is overfitting to noise in their ratings?
Number of stars on Github is not a great metric of the health of a project. TensorFlow, in particular, has received a lot of attention, so a lot of people have starred it, but this does not measure the number of non-Google contributors or what would happen to the project if Google abandoned it.
Note that I'm not saying TensorFlow is going to fail in the future; I'm saying that the measures used in this article are dubious at best.
Does nobody remember Elefant, Alex Smola's machine learning library that had a lot of hype then suddenly died after he left NICTA? The machine learning world is littered with tons of dead projects, and many of these died without a lot of warning. I'm concerned that this effect is only going to get worse now that machine learning libraries are often company-controlled instead of academic.
Hi there, I wrote the article. We also looked at data on things like number of forks and contributors on GitHub as well as massive rise in questions on StackOverflow (thanks to Delip Rao's great post here http://deliprao.com/archives/168 and invaluable ecosystem analysis from Francois Chollet). These other things were cut for purposes of length and because we felt a mainstream reader (which is more Bloomberg's audience) would be able to understand the 'stars' figure most easily of all metrics.
Obviously I agree with you that this measure isn't definitive, but we felt it was relatively easy to understand. And the rest of the reporting we did for this story bore out the idea that TF has gathered an unusual level of both enthusiasm and commitment in a short amount of time.
good points, though I would argue that tensorflow fundamentally is a way to define transformations. Think of it like assembly language. There are higher languages that expose higher level constructs (see keras)
Agreed it's not a great metric, but it is a good proxy especially if you suplement it with another source, such as the number of HN front page articles discussing TensorFlow. Given those two signals, I imagine Tensorflow will continue to gain momentum, at least in the short term.
However, like all SDKs (is that still a thing?), it can be easily be supplanted when something that is easier to use, fixes the flaws of, and is better documented comes out. This is doubly true given that the AI space is still (imo) in its infancy, wrt to developer tooling.
Politicians quote the number of "Likes" they get on social media, too. Stars are no different. It's a measure of interest, where the error bars are quite significant.
Hey, not all outdoor cats are pests! My cat is too incompetent to catch any birds and she just grazes on grass, so she's actually improving my yard by making it so I don't need to mow. Now if only I could get her to consistently eat it all to a certain height...
