Yann LeCun on a MIT Tech Review article that is all hype | Hacker News Reader

Yann LeCun on a MIT Tech Review article that is all hype

facebook.com

140 pointsbmease20 comments

This is a great response and I wish more scientists held publications accountable to their reporting. MIT Tech Review, Popular Science and Mechanics are read by a lot of people -- both literate in the your discipline and not, it's worthwhile to point the more nuanced view than what a 1000 word magazine article can point out. I've personally found reporting by ArsTechnica on recent security issues a good model. Wired occasionally comes through but has similar problems like this. Nautilus has its own biases but in general is good. I don't think bad reporting in these cases is necessarily out of malice but the lack of background on reporter's side on your field. And, perhaps sheer laziness. Remember when you procrastinated on writing that long overdue paper, I'd imagine reporters aren't immune to that too ;)

Convolutional Neural Nets are getting to a hype-level that I find pretty scary. We don't want another AI winter because people expect way too much too early without understanding the domain, only to lead to receeded interest in the field. Honest evaluation and crediting is invaluable to ensuring that.

Also --> What if there were a "rapgenius" for paper/article reviews where these comments from trusted sources can be curated and commented on? Not sure about viability, etc.. but could be interesting.

Genius is currently building a tool to achieve exactly this. It's in beta as we speak, just put genius.com/ before any article URL to see an annotated version of it. For eaxmple, here's an annotated version of a "Eloquent Javascript" chapter [1]. If you want beta access get an account and shoot me a message there [2].

We* have been annotating tech articles/papers for a while though, they were stored in a sub-sub-channel of Rap Genius though and didn't get as much exposure. The best example is teh analysis of NewsWeek's story "The Face Behind Bitcoin" [3].

[1] http://genius.com/eloquentjavascript.net/05_higher_order.htm...

[2] http://genius.com/fanahova

[3] http://genius.com/Leah-mcgrath-goodman-the-face-behind-bitco...

*I don't work at Genius, but I've been an editor/mod/intern for the past 2-3 years working on various things including Tech Genius.

Interesting, the analysis of NewsWeek story is pretty much what I was anticipating. It seems the Genius ecosystem has grown a fair amount since last I looked at it. I'd definitely like an account. Messaged you on Genius.

That's very cool. If people make a habit of using this, this could be really handy for annotating programming tutorials. Just modified the HN bookmarklet to make a new bookmarklet to reload the current page in Genius:

http://acjay.github.io/genius_bookmarklet.html

Haha great minds think alike! We've got a bookmarklet already: http://genius.com/bookmarklet (you can only see the page if you're a beta tester) and a chrome extension coming that will tell you when you browse to a page that's already been annotated and let you switch to the annotated version. It's pretty cool stuff!

Surprisingly or not, Marc Andreessen did a bit of annotating on that Bitcoin article.

Well reddit/HN has exactly this function for me. Sometimes I google the article title to get a reddit discussion. Specially in the more specialized subreddits (e.g. ComputerVision, Crypto, Mathematics) the discussion is very good.

See http://hypothes.is/ for an open-source annotation-tools project with a lot of big minds behind it.

Oh wow, this is actually perfect if it takes off - open standards, and straightforward from what I can tell.

MIT Tech Review sort of has a reputation for doing this.

Yes, they do. It should be embarrassing for MIT. They're not run by MIT. They're a startup in San Francisco which somehow acquired the rights to MIT's old in-house magazine. Their self-description: "an innovative, digitally oriented global media company whose reach is rapidly expanding."

Their articles about "nanotechnology" (they mean surface chemistry) are even worse.

They have gotten more dissociated from MIT in the past decade, but I believe MIT still owns the company that currently publishes it, so ultimately has some kind of control (or could in principle have control, if they cared to exercise it). I think overall they probably aren't unhappy with it, since MIT has long had quite a bit of tolerance (or even love) for the scientific hype machine, and the backlash tends to be infrequent and almost exclusively among academics, while the general public eats it up.

I don't really single out MIT; plenty of universities' PR offices, if given the chance, would publish a magazine similar or worse. Though perhaps with their large reach and stacks of cash, MIT could afford to be more careful than most, and set a better example.

I don't really single out MIT; plenty of universities' PR offices, if given the chance, would publish a magazine similar or worse.

cough Harvard Business Review cough

That seems like a stretch - according to Wikipedia : "The magazine is published by Technology Review, Inc, an independent media company owned by MIT. MIT's website lists it as an MIT publication, and the MIT News Office states that "the magazine often uses MIT expertise for some of its content."

They apparently have an office in SF these days but they also have an office in Cambridge. They're also still the MIT Alumni magazine. Honestly, they get wrapped up in the hype for new technologies sometimes but they seem a pretty reliable source for technology news overall. They are technology optimists and are in the business of writing about new technologies but it's easy to cherry-pick articles that get overly enthusiastic. By contrast, a recent cover story delved into troll hunting for example.

More interesting than a debunking of yet another hype linkbait article from MIT Tech Review, is that someone posted on Facebook a citation of his own 22-year old prior art.

Something about the Facebook backdrop makes that academic smackdown all the more special.

FYI, he works for Facebook. It'd be weird if he posted it to Google+ ;-)

Not really -- he's been an active Google+ poster even before he joined FB (https://plus.google.com/+YannLeCunPhD/posts)

Shameless tangential plug: Nicolas Vasilache from The Facebook AI Research group (Yann LeCun's team) is speaking next month at http://MLconf.com in NYC. We will be livestreaming the event for free. Watch @mlconf on 3/27 for the livestream link. If you would like to attend in person use discount code MLML for a discount.

*FULL DISCLOSURE: This is my event

" MIT Tech Review, Popular Science and Mechanics are read by a lot of people"

And the audience for all of them tends to want form over substance. That is why the quality in those magazines is so abyssmal.

Reply on news.ycombinator.com