Slashdot Mirror
Machine Learning Expert Michael Jordan On the Delusions of Big Data
First time accepted submitter agent elevator writes In a wide-ranging interview at IEEE Spectrum, Michael I. Jordan skewers a bunch of sacred cows, basically saying that: The overeager adoption of big data is likely to result in catastrophes of analysis comparable to a national epidemic of collapsing bridges. Hardware designers creating chips based on the human brain are engaged in a faith-based undertaking likely to prove a fool's errand; and despite recent claims to the contrary, we are no further along with computer vision than we were with physics when Isaac Newton sat under his apple tree.
A man of many talents.
... and despite recent claims to the contrary, we are no further along with computer vision than we were with physics when Isaac Newton sat under his apple tree.
That's true, I looked into object recognition for image classification by content. Face recognition is proceeding fairly nicely but doing stuff like just programmatically classifying/tagging images by whether they contain a car, airplane, house, tree, dog, mountain .... without even trying to do things like identifying the type of airplane/dog/car is pretty much undoable in any reasonable amount of time with human level accuracy needed on garden variety PCs and tablets which is the application I'd be interested in. The fastest and most accurate image classifier/tagger is still a human. Am still looking forward to they day that changes but I'm not sure that will be within my lifetime.
Only to idiots, are orders laws.
-- Henning von Tresckow
This is why I don't take Ray Kurzweil's predictions seriously. People like Prof. Jordan, who would actually make the vision become reality, dont take Kurzweil's ideas seriously.
There's plenty of reasons I can think of why I'd prefer image recognition on my phone rather than the cloud. Privacy, for one. If you let FB tag your photos with the names of the people in it (after teaching it those names), what do you think happens to that data? You might not even want to share the photo or video stream with anyone... Another reason is that we still do not live in a world with ubiquitous and cheap mobile data. Travel abroad, and you'll find out quickly why cloud-based services like Waze aren't always a viable option.
If construction was anything like programming, an incorrectly fitted lock would bring down the entire building...
As it happens, I am a computer vision expert.
I do wonder how much useful stuff was done with the results from physics back then as opposed to emperical hand-hacking of everything. I suspect not much.
Computer vision has a long way to go. On the other hand, there are plenty of things which it does do, some of which are more or less impossible otherwise.
OCR is very useful. It runs the mail system of many countries and has plenty of use when it comes to digitising old documents. This would be possible, but deeply tedious by hand.
Structure from motion is used heavily in the film industry to work out 3D structure and motion for placing virtual objects. Almost impossible to do well without computer vision.
Photo stitching for automatic panoramas. Classic CV system, and my phone comes with it built in.
Number plate recognition. Apart from the rather unpleasant big brother potential, London's congestion charging system runs off this and it does very good things for London.
Those cameras/phones with face detection built in. Not sure how useful it is but it works.
Lego Fusion is a recently released game which appears to rely on computer vision.
Oh those phone based barcode and QR scanners. Very useful.
The pick and place machines which use vision for accurate placement.
This machine which is really awesome: https://www.youtube.com/watch?...
Lots of other industrial things are controlled by CV.
Certain types of super resolution microscopy are based on computer vision.
And that's just a few off the top of my head.
So yeah computer vision has a long way to go. On the other hand, it's out there doing real things right now. It might not be very advanced CV (the industrial stuff often is not because it needs to be reliable), but it's still CV and it's still being used.
SJW n. One who posts facts.
funnier written as:
.
.
.
.
.
Latency.
Of course there's a reason, and it's not privacy or latency, it's cost. If you want to provide a free app that does image recognition, who pays for the cloud servers? The user has horsepower on his phone that rivals an early Cray. Only an idiot would go to the trouble and expense of putting this into the cloud, only to end up paying for the cloud service while struggling to get any upstream revenue. I notice that the two examples you give are from companies that run cloud systems themselves. Nobody does this when they don't own the cloud, and/or are trying to promote cloud computing as a thing. So no, people do not need to stop thinking this way. People need to stop looking at "the cloud" as some kind of magical panacea for not-solving efficiency problems.
I am doing a postdoc in applied statistics/machine learning and I was very surprised by this interview since it is contradictory to what Michael Jordan has himself expressed as an invited speaker at conferences as well as what his most recent research projects are focused at. It appears that, according to Michael Jordan himself as expressed on his webpage, the article is a hack-job where the journalist is completely misrepresenting his view on big data. To quote:
I’ve found myself engaged with the Media recently (...) for an interview that has been published in the IEEE Spectrum.
That latter process was disillusioning. Well, perhaps a better way to say it is that I didn’t harbor that many illusions about science and technology journalism going in, and the process left me with even fewer.
The interview is here: http://spectrum.ieee.org/robotics/artificial-intelligence/machinelearning-maestro-michael-jordan-on-the-delusions-of-big-data-and-other-huge-engineering-efforts
Read the title and the first paragraph and attempt to infer what’s in the body of the interview. Now go read the interview and see what you think about the choice of title.
The title contains the phrase “The Delusions of Big Data and Other Huge Engineering Efforts”. It took me a moment to realize that this was the title that had been placed (without my knowledge) on the interview I did a couple of weeks ago. Anyway who knows me, or who’s attended any of my recent talks knows that I don’t feel that Big Data is a delusion at all; rather, it’s a transformative topic, one that is changing academia (e.g., for the first time in my 25-year career, a topic has emerged that almost everyone in academia feels is on the critical path for their sub-discipline), and is changing society (most notably, the micro-economies made possible by learning about individual preferences and then connecting suppliers and consumers directly are transformative). But most of all, from my point of view, it’s a *major engineering and mathematical challenge*, one that will not be solved by just gluing together a few existing ideas from statistics, optimization, databases and computer systems.
Source: https://amplab.cs.berkeley.edu/2014/10/22/big-data-hype-the-media-and-other-provocative-words-to-put-in-a-title/
No, seriously. Here are some choice quotes:
"I read all the time about engineers describing their new chip designs in what seems to me to be an incredible abuse of language. They talk about the “neurons” or the “synapses” on their chips. But that can’t possibly be the case; a neuron is a living, breathing cell of unbelievable complexity."
"It’s always been my impression that when people in computer science describe how the brain works, they are making horribly reductionist statements that you would never hear from neuroscientists."
"Lately there seems to be an epidemic of stories about how computers have tackled the vision problem, and that computers have become just as good as people at vision."
"Even in facial recognition, my impression is that it still only works if you’ve got pretty clean images to begin with."
"I have a hobby of searching for information about silly Kickstarter projects, mostly to see how preposterous they are, and I end up getting served ads from the same companies for many months."
Here's the catch: all of these quotes are from the interviewer. Jordan has a lot of really nuanced claims here, but it's clear that the interviewer has an agenda of his own.
That's selective quoting taken to the extreme. The GP was talking about the applications he'd be interested in. Do you know what he's interested in? I don't. But I do have a friend who escapes the Scottish winter every year to go searching for undiscovered orchid specieses in a Vietnamese rainforest. Now call me a pessimist, but I doubt he's going to get a 3G signal out there. What if he wants to check if a flower is a known species? He can do that within his area (the orchids) but he can't be expected to have an encyclopedic knowledge of all extant plant-life. Wouldn't it be nice if his mobile phone could flag up a potentially unknown species that he stumbles across, giving him to opportunity to take a sample back for analysis?
Or a less extreme example -- if I'm travelling, I want my translation app to work even when I can't get an internet connection.
But more to the point, your message takes for granted the problem that TFA alludes to: when you say any even moderately heavy compute job is shipped off to the cloud it accepts AI-type tasks as being computationally complex, but that is due to the lack of progress within the field. We're still effectively "brute-forcing" the problem in many ways, and instead of looking for better algorithms to handle the process, we're just scaling up the same process, running it on "big iron" and calling it progress because we can handle fancier-looking pictures.
Got them moderator blues I blieve I walk out the do', With these mod-points I been gettin', I 'most never post no mo'
As someone who was involved in the previous neural network hype cycle (late 80s, early 90s), I'd have to agree with him that we go through these cycles, where a particular approach gain ascendency, then is shown to not work as well as the hype, and then gets rejected. On the inside, however, lots of good work continues to be done. The press (and then in popular opinion) keeps saying 'this is it, we're really close to AI' or somethign similar, and then when it doesn't pan out, then it is considered a bust. But, we are making progress, we know more than we did last year, and a lot more than 10 years ago. It is just that the problem is hard, and we're still trying to figure out some basic principles, so don't expect us to be there yet.
The more people I meet, the better I like my dog.
Actually, I believe he is not even a hack in the sense that he never had done any real technology connected to his ravings. He is a delusional loon that is completely disconnected from reality.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.