How Robots and Algorithms Are Taking Over | Hacker News Reader

How Robots and Algorithms Are Taking Over

nybooks.com

59 pointsjonathansizz23 comments

"... In fact, this is already happening, in part because programmers increasingly rely on 'self-correcting' code—that is, code that debugs and rewrites itself ..."

What on earth is the author talking about? I've never heard of self-correcting code, certainly never used it, and can't find a reasonable reference to it with a google search. Is he mistakenly referring to "Error-Correcting Code", that is, using redundancy in data storage or transmission to catch and correct for random bit flips?

I can't believe the hubris of the tech industry. I bet no one even considers the danger that their ECC RAM becomes self-aware.

Hype article with FUD

Wow, my opinion of the editing process at the New York Review of Books has just dropped, by a lot. It's not just that it's a bizarre claim, but that it's also stated so matter-of-factly. Honestly, the writer probably talked to a coder friend who uses an IDE, and the way the coder described auto-completion and type-checking extensions sounded very much like "self-correcting" code.

No matter what you think of difficulties of the Turing halting problem, I'd say that writers depend --and can benefit -- far more on "self-correcting" writing (particularly auto-correct) than programmers do on self-correcting code.

I think you're overreacting. The author could well be talking about neural nets, deep learning, etc. We may choose to nitpick or discuss the overall point.

When anyone critiques information technology who is not already in the community, we seem to prefer to nitpick.

That could be the next part "machines that are able to learn on the job", but that's not "code that debugs and rewrites itself".

It's not a far stretch that laymen conflate "code" with "statistical learning algorithms."

I googled 'self-correcting code' because I thought the same thing! Definitely a bizarre claim.

Perhaps linters, auto-formatters, auto-complete, etc. I could understand that it could look like "self-correcting code" to a non-programmer.

Analogously , "journalists rely on 'self-correcting' articles - that is, an article that 'debugs' and rewrites itself...".

http://www.cs.cmu.edu/~clegoues/docs/legoues-icse09.pdf "Automatically Finding Patches Using Genetic Programming ... We introduce a fully automated method for locating and repairing bugs in software" http://dijkstra.cs.virginia.edu/genprog/

Though I think the author was talking more about coding up a neural net, where it can be said that the net learns the features itself, and optimizes a loss function, without much reliance on a superviser.

Perhaps he's talking about compilers optimizing code?

Given the level of cluelesness needed to write something like this, I'd guess that he's referring to smart Java and C# IDEs. To somebody clueless, all this automatic refactoring, code cleanup suggestions, stylistic fixes, etc. could look like "code that debugs and rewrites itself".

It's a hodge-podge of semi-accurate statements about very minor trends (minor in employment terms) and inaccurate statements that it's hard to critique. Check this partial paragraph with my thoughts as I read it:

"Meanwhile, algorithms are writing most corporate reports [What? what kind of reports are we talking about? At first I pictured those glossy end-of-year things for analysts, but maybe not, anyway - citation needed!] analyzing intelligence data for the NSA and CIA, [Analyzing? Or just winnowing hopefully significant items from oceans of chaff for real analysts to look at?] reading mammograms [OK, valid hit], grading tests [then why are teachers and professors always still bitching about the time they spend grading?] , and sniffing out plagiarism [valid, but only a minor adjunct to a professor's grading life]. Computers fly planes—Nicholas Carr points out that the average airline pilot is now at the helm of an airplane for about three minutes per flight [False impression, there has to be an awake and alert trained pilot at "the helm" 100% of the time regardless of the autopilot]—and they compose music [BAD music that nobody but a few academics ever listen to and that certainly does not sell] and pick which pop songs should be recorded based on which chord progressions and riffs were hits in the past [? that would explain a lot about current pop...]. Computers pursue drug development—a robot in the UK named Eve may have just found a new compound to treat malaria [yes there are interesting machines that try thousands of molecules to find ones that might work, but that in no way displaces any human bodies; these are NEW discoveries that wouldn't be made otherwise]—and fill pharmacy vials [as pick-and-pack machinery has been doing for decades]...

This article is a mess.

Your comment is a mess. Most of your "criticisms" seem to follow the logic that if you haven't personally seen it before, it doesn't exist.

Programming will probably look entirely different from what it looks like now in two decades, but I doubt it will be entirely replaced as a profession in that time. If it has, the singularity will have been started and everything will be entirely different anyways.

I think it's pretty foolish that most programmers assume they will always be "working above the api".

http://rein.pk/replacing-middle-management-with-apis/

edit: To clarify, I think AI is the wrong question it's a question of the layers of abstraction and where it puts you in the labor force. Who cares what inhuman force is telling you what to do or how fast to do it?

I assume there will be no clear definition of where the API begins and where it ends when this happens. The various systems which employ humans will probably be mixture of AI and humans working together through multiple interfaces.

I have a hunch the author read someone's explanation of "machine learning algorithm" and the author changed it to "self-correcting code" along the way.

This is supposed to be a book review, but seems to be mostly uninformed blithering by the reviewer.

Actual physical robots still aren't very good at dealing with the real world and cost too much. Progress is being made, but it's slow. Having been in the field, it's been discouraging how slow. Robot manipulation in unstructured environments is still very weak. "Pick up thing and put it somewhere" works reasonably well. Beyond that, performance is poor.

The robot cost problem is severe. Look at prices on industrial robots or intelligent farming equipment.

On the other hand, jobs which involve sitting at a desk and dealing with inputs and outputs which come in via wires (or paper, for retro outfits) are looking vulnerable. If they haven't been automated already. Computers are just so cheap now.

This leads to "Machines should think, people should work." Computers are better at organization and scheduling than humans. Humans are better at the grunt work, for now. Consider Uber. Or Amazon's order-fulfillment operations. It's not clear how far that concept will reach. HouseCall tried to do it for plumbers and repair services, but didn't get much traction. Uber is in a space where the product is uniform and the pricing model is simple. HouseCall is not.

After reading a review like this, it occurs to me that it's probably not too hard to write a program to review popular non-fiction books. Grind through the text. Run each chapter through a summarizer like the one that used to come with Microsoft Word. From the summary, find key phrases and look them up with Google. Find popular comments on the same subject from people with some notability. Look at the Amazon sales statistics and generate some prose using a modified business earnings story generator. Generate a clickbait title and feed into the Demand Media or AOL content mills.

Articles like this, which I come across enough to have mental category for them, are hard to take seriously.

It seems to me that of the great problems facing the world that that this is not one of them. As opposed to vague hypothetical troubles down the road, there is no shortage of vexing problems clamoring for attention right now.

I think improvements to our political and economic systems, which are under direct human control, would pay big dividends.

To the extent automation is a malevolent force for humanity, it seems far more likely that the people and institutions who control this technology will be the ones who decide whether it is used for good or ill.

"While these machines cannot think, per se, ..."

The author appears to have no idea that this might be a controversial statement.

Reply on news.ycombinator.com