If you think this is a symptom of AI winter, then you are probably wasting time on outdated/dysfunctional models or models that aren't suited for what you want to accomplish. Looking e.g. at Google Duplex (better voice synchronization than Vocaloid I use for making music), this pushed state-of-art to unbelievable levels in hard-to-address domains. I believe the whole SW industry will be living next 10 years from gradual addition of these concepts into production.
If you think Deep (Reinforcement) Learning is going to solve AGI, you are out of luck. If you however think it's useless and won't bring us anywhere, you are guaranteed to be wrong. Frankly, if you are daily working with Deep Learning, you are probably not seeing the big picture (i.e. how horrible methods used in real-life are and how you can easily get very economical 5% benefit of just plugging in Deep Learning somewhere in the pipeline; this might seem little but managers would kill for 5% of extra profit).
Understand that in pop-sci circles over the past several years the general public is being exposed to stories warning about the singularity by well respected people like Stephen Hawking and Elon Musk (http://time.com/3614349/artificial-intelligence-singularity-...). Autonomous vehicles are on the roads and Boston Dynamics is showing very real robot demonstrations. Deep learning is breaking records in what we thought was possible with machine learning. All of this progress has excited an irrational exuberance in the general public.
But people don't have a good concept of what these technologies can't do, mainly because researchers, business people, and journalists don't want to tell them--they want the money and attention. But eventually the general public wises up to the unfulfillment of expectations, and drives their attention elsewhere. Here we have the AI winter.
Further, I have repeatedly heard people who should know better, with very fancy advanced degrees, chant variants of "Deep Learning gets better with more data" and/or "Deep Learning makes feature engineering obsolete" as if they are trying to convince everyone around them as well as themselves that these two fallacious assumptions are the revealed truth handed down to mere mortals by the 4 horsemen of the field.
That said, if you put your ~10,000 hours into this, and keep up with the field, it's pretty impressive what high-dimensional classification and regression can do. Judea Pearl concurs: https://www.theatlantic.com/technology/archive/2018/05/machi...
My personal (and admittedly biased) belief is that if you combine DL with GOFAI and/or simulation, you can indeed work magic. AlphaZero is strong evidence of that, no? And the author of the article in this thread is apparently attempting to do the same sort of thing for self-driving cars. I wouldn't call this part of the field irrational exuberance, I'd call it amazing.
I think even if you avoid constructing features, you are basically doing a similar process where a single change in a hyper-parameter can have significant effects:
- internal structure of a model (what types of blocks are you using and how do you connect them, what are they capable of together, how do gradients propagate?)
- loss function (great results come only if you use a fitting loss function)
- category weights (i.e. improving under-represented classes)
- image/data augmentation (self-driving car won't work without significant augmentation at all)
- properly set-up optimizer
The good thing here is that you can automate optimization of these to a large extent if you have a cluster of machines and a way to orchestrate meta-optimization of slightly changed models. With feature engineering you just have to do all the work upfront, thinking what might be important, and often you just miss important parts of features :-(
curious, if there is any good quality open source project for this..
And more importantly, business and government leaders wise up and turn off the money tap.
I think they also happen when the best ideas in the field run into the brick wall of insufficiently developed computer technology. I remember writing code for a perceptron in the '90s on an 8 bit system, 64 k RAM - it's laughable.
But right now compute power and data storage seem plentiful, so rumors of the current wave's demise appear exaggerated.
Most of the software world will have to move on stuff like Haskell or functional language. As of now bulk(almost all) of our people are trained to program in C based languages.
It won't be easy. There will be a renewal for high demand software jobs.
A symptom of capitalism and marketing trying to push shit they don't understand
- how good humans are in detecting cancer (hint: not very good) and if having an automated system even as a "second opinion" next to an expert might not be useful?
- there are metrics for capturing true/false positives/negatives one can focus on during learning optimization
From studies you might have noticed that expert radiologists have e.g. F1-score at 0.45 and on average they score 0.39, which sounds really bad. Your system manages to push average to 0.44, which might be worse than the best radiologist out there, but better than an average radiologist . Is this really being oversold? (I am not addressing possible problems with overly optimistic datasets etc. which are real concerns)
The problem AI runs into is that with too much faith in the machine, people STOP thinking and believe the machine. Where you might get a .44 detection rate on radiology data alone, that radiologist with a .39 or a doctor can consult alternate streams of information. The AI may still be helpful in reinforcing a decision to continue scrutinizing a set of problem.
AI's as we call them today are better referred to as expert systems. AI carries too much baggage to be thrown around Willy nilly. An expert system may beat out a human at interpreting large unintuitive datasets, but they aren't generally testable, and like it or not, it will remain a tough sell in any situation where lives are on the line.
I'm not saying it isn't worth researching, but AI will continue to fight an uphill battle in terms of public acceptance outside of research or analytics spaces, and overselling or being anything but straightforward about what is going on under the hood will NOT help.
In medicine, I want everyone to apply appropriate skepticism to important results, and I don't want to enable lazy radiologists to zone out and press 'Y' all day. I want all the doctors to be maximally mentally engaged. Skepticism of an incorrect radiologist report recently saved my dad from some dangerous, and in his case unnecessary, treatment.
And there's probably not even a boolean "ground truth" in complicated bio-medicine problems. Sometimes the right call is neither yes or no, but: this is not like anything I've seen before, I can't give a decision either way, I need further tests.
State of the art in numbers:
Image Classification - ~$55, 9hrs (ImageNet)
Object Detection - ~$40, 6hrs (COCO)
Machine Translation - ~$40, 6hrs (WMT '14 EN-DE)
Question Answering - ~$5, 0.8hrs (SQuAD)
Speech recognition - ~$90, 13hrs (LibriSpeech)
Language Modeling - ~$490, 74hrs (LM1B)
I don't know. Duplex equipped with a way to minimize his own uncertainties sounds quite scary.
Microsoft OTOH quietly shipped the equivalent in China last month: https://www.theverge.com/2018/5/22/17379508/microsoft-xiaoic...
Google has lost a lot of steam lately IMO. Facebook is releasing better tools and Microsoft, the company they nearly vanquished a decade ago, is releasing better products. Google does remain the master of its own hype though.
Google nearly vanquished Microsoft a decade ago?
Where can I read more about this bit of history :) ?
IMO, Axios  seem to do a better job of criticizing Google's Duplex AI claims, as they repeatedly reached out to their contacts at Google for answers.
Microsoft was previously the gatekeeper to almost every interaction with software (roughly 1992 - 2002). I don't know of good books on it but Tim O'Reilly wrote quite a bit about Web 2.0.
I'm quite familiar with Google's history and would not characterize them as having vanquished Microsoft.
For the most part, Microsoft doesn't need to lose for Google to win (except of course in the realm of web search and office productivity).
Unfortunately, by the time of my brief stint at Google, the place was a professional dead-end where most of the hirees got smoke blown up their patooties at orientation about how amazing they were to be accepted into Google, only to be blind allocated into me-too MVPs of stuff they'd read about on TechCrunch. All IMO of course.
That said, I met the early Google Brain team there and I apparently made a sufficiently negative first impression for one of their leaders to hold a grudge against me 6 years later, explaining at last who it was that had blacklisted me there. So at least that mystery is solved.
PS It was pretty obvious these were voice actors in a studio conversing with the AI. That is impressive, but speaking as a former DJ myself, when one has any degree of voice training, one pronounces words without much accent and without slurring them together. Google will likely never admit anything here: they don't have to.
But I will give Alphabet a point for Waymo being the most professionally-responsible self-driving car effort so far. Compare and contrast with Tesla and Uber.
We haven't found life outside this planet, and we haven't created life in a lab, therefore n=1 for assessing probability of life outside earth (which means we can't calculate a probability for this yet). Likewise, we haven't created anything remotely like animal intelligence (let alone human) and we have no good theory regarding how it works, so n=1 for existing forms of general intelligence.
Note that I'm not saying there can be no extraterrestrial life or that we will never develop AGI, just that I haven't seen any evidence at this point in time that any opinions for or against their possibility are anything more than baseless speculation.
"To train the system in a new domain, we use real-time supervised training. This is comparable to the training practices of many disciplines, where an instructor supervises a student as they are doing their job, providing guidance as needed, and making sure that the task is performed at the instructor’s level of quality. In the Duplex system, experienced operators act as the instructors. By monitoring the system as it makes phone calls in a new domain, they can affect the behavior of the system in real time as needed. This continues until the system performs at the desired quality level, at which point the supervision stops and the system can make calls autonomously." --
OK, but 83% ROC/AUC is nothing to be bragging about. ROC/AUC routinely overstates the performance of a classifier anyway, and even so, ~80% values aren't that great in any domain. I wouldn't trust my life to that level of performance, unless I had no other choice.
You're basically making the author's case: deep learning clearly outperforms on certain classes of problems, and easily "generalizes" to modest performance on lots of others. But leaping from that to "radiology robots are almost here!" is folly.
There is also higher chance that next state-of-art model would push it significantly over 83% or best human radiologist at some point in the future, so it might not be very economical to train humans to become even better (i.e. dedicate your life to focus on radiology diagnostics only).
The main critique for CheXNet I've read was focused on the NIH dataset itself, not the model. The model generalizes quite well across multiple visual domains, given proper augmentation.
Except they don't. See the table in the original post. Also, comparing the "average" radiologist by F1 scores from a single experiment (as you've done in other comments here) is meaningless.
Unless my doctor is exactly average (and isn't incorporating additional information, or smart enough to be optimizing for false positive/negative rates relative to cost), comparison to average statistics is academic. But I don't really need to tell you this -- your comment has so many caveats that you're clearly already aware of the limits of the method.
On one hand, we have one commenter saying he can train a model to do a specific thing with a specific quantitative metric, to demonstrate how deep learning can incredibly powerful/useful.
On the other hand, we have another commenter saying "But this won't replace my doctor!" and therefore deep learning is overhyped.
The two sides aren't even talking about the same thing.
That kind of hyperventilating stuff is easy to brush off. The problem with deep-learning hype is that comments like "my classifier gets a ROC/AUC score of 0.8 with barely any work!" are presented as meaningful. The difference between a 0.8 AUC and a usable medical technology means that most of the work is ahead of you.
So I think it's come down to conflict between
1. Which the author is trying to present
2. What an astute reader might interpret it as
3. What an astute reader might worry an uninformed reader might interpret it as
And my feeling is that, given all the talk about hype in pop-sci, we're actually on point 3 now, even when the author and reader are actually talking about something reasonable. Whereas personally I'm more interested in the research and interpretations from experts, which I find tend to be not so problematic.
Just to get back to this point: what if the vision system of your doctor is below average and you augment her by giving her a statistically better vision system, while allowing her to use the additional sources as she sees fit. Wouldn't be that an improvement? We are talking about vision subsystem here, not the whole "reasoning package" human doctors posses.
On just about every test set, the model is beaten by radiologists. Even the mean performance is underwhelming.
In their paper they even used "weaker" DenseNet-121 instead of DenseNet-169 for Mura/bones. DenseNet-BC I tried is another refinement of the same approach.
Basically, I see this as "everyone sucks, but the AI maybe sucks a little less than the worst of our radiologists, on average"
Some people mention Matthews correlation coefficients, Youden's J statistic, Cohen's kappa etc. but I haven't seen them in any Deep Learning paper so far and I bet they have large blindspots as well.
Of course! Using DenseNet-BC-100-12 to increase ROC AUC, it was so obvious!
The author is clearly informed and takes a strong, historical view of the situation. Looking at what the really smart people who brought us this innovation have said and done lately is a good start imo (just one datum of course, but there are others in this interesting survey).
Deepmind hasn't shown anything breathtaking since their Alpha Go zero.
Another thing to consider about Alpha Go and Alpha Go Zero is the vast, vast amount of computing firepower that this application mobilized. While it was often repeated that ordinary Go program weren't making progress, this wasn't true - the best, amateur programs had gotten to about 2 Dan amateur using Makov Tree Search. Alpha Go added CNNs for it's weighting function and petabytes of power for it's process and got effectiveness up to best in the world, 9 Dan professional, (maybe 11 Dan amateur for pure comparison). 
Alpha Go Zero was supposedly even more powerful, learned without human intervention. BUT it cost petabytes and petabytes of flops, expensive enough that they released a total of ten or twenty Alpha Go Zero game to the world, labeled "A great gift".
The author convenniently reproduces the chart of power versus results. Look at it, consider it. Consider the chart in the context of Moore's Law retreating. The problems of Alpha Zero generalizes as described in the article.
The author could also have dived into the troubling question as of "AI as ordinary computer application" (what does testing, debugging, interface design, etc mean when the app is automatically generated in an ad-hoc fashion) or "explainability". But when you can paint a troubling picture without these gnawing problems appearing, you've done well.
They went on to make AlphaZero, a generalised version that could learn chess, shogi or any similar game. The chess version beat a leading conventional chess program 28 wins, 0 losses, and 72 draws.
That seemed impressive to me.
Also they used loads of compute during the training but not so much during play.(5000 TPUs, 4TPUs).
Also it got better than humans in those games from scratch in about 4 hours whereas humans have had 2000 years to study them so you can forgive it some resource usage.
What was impressive was the way Stockfish9 was beaten. AlphaZero played like a human player, making sacrifices for position that stockfish thought were detrimental. When it played as white, the fact that is mostly started with the Queen pawn (despite that the King pawn is "best by test") and the way AlphaZero used Stockfish pawnstructure and tempo to basicaly remove a bishop from the game was magical.
Yes, since its a game, it's "useless", but it allowed me (and i'm not the only one) to be a bit better at chess. It's not world hunger, not climate change, it's just a bit of distraction for some people.
PS: I was part of the people thinking that Genetic algorithm+deep learning was not enough to emulate human logical capacities, AlphaZero vs Stockfish games made me admit i was wrong (even if i still think it only works inside well-defined environments)
Just because Fischer preferred 1. e4, it doesn't make it better than other openings.
Playing like a human for me also means making human mistakes. A chess-playing computer playing like a 4000 rated "human" is useless, one that can be configured to play at different ELOs is more interesting, although most can do that and there's no ML needed, nor huge amounts of computing power.
Without its opening database and without its endgame tablebase?
Frankly, the Stockfish vs AlphaZero match was the beginning of the AI Winter in my mind. The fact that they disabled Stockfish's primary databases was incredibly fishy IMO and is a major detriment to their paper.
Stockfish's engine is designed to only work in the midgame of Chess. Remove the opening database and remove the endgame database, and you're not really playing against Stockfish anymore.
The fact that Stockfish's opening was severely gimped is not a surprise to anybody in the Chess community. Stockfish didn't have its opening database enabled... for some reason.
Most games are also closed systems, and conveniently grokkable systems, with enumerable search spaces. Which gives us easily produceable measures of the contraptions' abilities.
Whether this is the most effective path to understanding deeper questions about intelligence is an open question.
But I don't think it's fair to say that deeper questions and problems are being foregone simply to play games.
I think most 'games researchers' are pursuing these paths because they themselves and no one else has put forth any other suggestion that makes them think, "hmm, that's a really good idea, that seems like it might be viable and there is probably something interesting we could learn from it."
Do you have any suggestions?
And comparing Alpha Go Zero against those "other chess programs that existed for 30 years" is exactly missing the point also.
Those programs were not constructed with zero-knowledge. They were carefully crafted by human players to achieve the result. Are we also going to count in all the brain processing power and the time spent by those researchers to learn to play chess? Alpha Go Zero did not need any of that, besides the knowledge about the basic rules of the game. Who compare compute requirements for 2 programs that have fundamentally different goals and achievements? One is carefully crafted by human intervention. The other one learns a new game without prior knowledge...
Sounds more like religion and less like science to me.
I guess we could argue until the end of the world that no intelligence will emerge from more and more clever ways of brute-forcing your way out of problems in a finite space with perfect information. But that's what I think.
On the topic of the different algorithmic approaches, I find it so fascinating how different these two approaches actually end up looking when analyzed by a professional commentator. When you watch the new style with a chess commentator, it feels a lot like listening to the analysis of a human game. The algorithm has very clearly captured strategic concepts in its neural network. Meanwhile, with older chess engines there is a tendency to get to positions where the computer clearly doesn't know what its doing. The game reaches a strategic point and the things its supposed to do are beyond the horizon of moves it can computer by brute force. So it plays stupid. These are the positions that, even now, human players can beat better than human old style chess engines at.
But attacking not-well-constrained problems is what's needed to show real progress in AI these days, right?
This. Learning to play a game is one thing. Learning how to teach computers to learn a game is another thing. Yes chess programs have been good before, but that's missing the point a little bit. The novel bit is not that it can beat another computer, but how it learned how to do so.
That's a pretty major shift for humanity.
But it's a mistake to think that a system learning by playing against itself is something new. Arthur Samuel's draughts (chequers) program did that in 1959.
It's not that it's new, it's that they've achieved it. Chess was orders of magnitude harder than draughts. The solution for draughts didn't scale to chess but Alpha Go zero showed that chess was ridiculously easy for it once it had learned Go.
Big Blue is fine - it's referring to the company and not the machine. From Wikipedia "Big Blue is a nickname for IBM"
Consider it as the perfect lab.
Seems like a lab so simplified that I'm unconvinced of its general applicability. Perfect knowledge of the situation and a very limited set of valid moves at any one time.
an awful lot of graph and optimization problems. See for instance some examples in https://en.wikipedia.org/wiki/A*_search_algorithm
Did they manage to extend it to games with hidden and imperfect information?
(Say, chess with fog of war also known as Dark Chess. Phantom Go. Pathfinding equivalent would be an incremental search.)
Edit: I see they are working on it, predictive state memory paper (MERLIN) is promising but not there yet.
(You said problems, not games...)
The thing is, an algorithm that can work with fewer samples and robustly tolerating mistakes in datasets (also known as imperfect information) will be vastly cheaper and easier to operate. Less tedious sample data collection and labelling.
Working with lacking and erroneous information (without known error value) is necessarily a crucial step towards AGI; as is extracting structure from such data.
This is the difference between an engineering problem and research problem.
I completely agree about the importance of imperfect information problems. In practice, many techniques handle some label noise, but not optimally. Even MNIST is much easier to solve if you remove the one incorrectly-labeled training example. (one! Which is barely noise. Though as a reassuring example from the classification domain, JFT is noisy and still results in better real world performance than just training on imagenet.)
I guess in the same way as lab chemistry isn't interesting anymore ? (Since it often happens in unrealistically clean equipment :-)
I think there is nothing preventing lab research from going on at the same time as industrialization of yesterday's results. Quite on the contrary: in the long run they often depend on each other.
The real challenge is to devise a general algorithm that will learn to be a good poker player in thousands of games, strategically, from just a bunch of games played. DeepStack AI required 10 million simulated games. Good human players outperform it at intermediate training stages.
And then the other part is figuring out actual rules of a harder game...
A good example of a game of imperfect information is poker, because players have a private hand which is known only to them. Whereas all possible future states of a chess game can be narrowed down according to the current game state, the fundamental uncertainty of poker means there is a combinatorial explosion involved in predicting future states. There's also the element of chance in poker, which further muddies the waters.
Board games are often (but not always) games of perfect and complete information. Card games are typically games of imperfect and complete information. This latter term, "complete information", means that even if not all of the game state is public, the intrinsic rules and structure of the game are public. Both chess and poker are complete, because we know the rules, win conditions and incentives for all players.
This is all to say that games of perfect information are relatively easy for a computer to win, while games of imperfect information are harder. And of course, games of incomplete information can be much more difficult :)
If this were true, there would be a vast demand for grandmasters in commerce, government, the military... and there just isn’t. Poker players suffer from similar delusions about how their game can be generalised to other domains.
Oh that's so true
Poker players in the real life would give up more often than not, whenever they didn't know enough about a situation or they didn't have enough resources for a win with a high probability.
And people can call your bluff even if you fold.
I suspect that chess as a metagame is just so far developed that being "good at chess" means your general ability is really overtrained for chess.
Can AI make the world better? It can, but it won't since we are humans, and humans will weaponize technology every chance it gets. Of course some positive uses will come, but the negative ones will be incredibly destructive.
The practical uses of these technologies don't always make national news.
I'm sure you would also have scoffed at the "pointless impractical, wasteful use of our brightest minds" to make the the Flyer hang in the air for 30 yards at Kitty Hawk.
We solved nothing.
IBM Deep Blue doesn't exactly think like humans do.
Most of our algorithms really are 'better brute force'.
Side observers are taking joy in the risker plays that it did -- reminded them of certain grand-masters I suppose -- but that still doesn't mean AGZ is close to any form of intelligence at all. Those "riskier moves" are probably just a way to more quickly reduce the problem space anyway.
It seriously reminds me more and more of religion, the AI area these days.
Most humans don't live 2000 years. And realistically don't spend that much of their time or computing power on studying chess. Surely a computer can be more focused at this and the 4h are impressive. But this comparison seems flawed to me.
Remember people reach peak play in ~15 years, but they don't nessisarily keep up with advances.
PS: You see this across a huge range of fields from running, figure skating, to music people simply spend more time and resources getting better.
So, in this sense, it's kind of like taking a human, teaching them the exact rules of the game and showing them how to run calculations, and then telling them to sit in a room playing games against themselves. In my experience from chess, you'd be at a huge disadvantage if you started with this zero-knowledge handicap.
One problem is that we can't play millions of games against ourselves in a few hours. We can play a few games, grow tired, and then need to go do something else. Come back the next day, repeat. It's a very slow process, and we have to worry about other things in life. How much of one's time and focus can be used on learning a game? You could spend 12 hours a day, if you had no other responsibilities, I guess. That might be counter productive, though. We just don't have the same capacity.
If you artificially limited AlphaGo to human capacity, then my money would be on the human being a superior player.
In a not equal fight, and the results are still not published. I'm not claiming that AlphaZero wouldn't win, but that test was pure garbage.
I agree AlphaZero had fancier hardware and so it wasn't really a fair fight.
Few would care. Your examiner doesn't give you extra marks on a given problem for finishing your homework quickly.
Just because alpha zero doesn't solve the problem you want it to doesn't mean that advancements aren't being made that matter to someone else. To ignore that seems disingenuous.
If you want an idea of where machine learning is in the scheme of things, the best thing to do is listen to the experts. _None_ of them have promised wild general intelligence any time soon. All of them have said "this is just the beginning, it's a long process." Science is incremental and machine learning is no different in that regard.
You'll continue to see incremental progress in the field, with occasional demonstrations and applications that make you go "wow". But most of the advances will be of interest to academics, not the general public. That in no way makes them less valuable.
The field of ML/AI produces useful technologies with many real applications. Funding for this basic science isn't going away. The media will eventually tire of the AI hype once the "wow" factor of these new technologies wears off. Maybe the goal posts will move again and suddenly all the current technology won't be called "AI" anymore, but it will still be funded and the science will still advance.
It's not the exciting prediction you were looking for I'm sure, but a boring realistic one.
What make this 3rd/4th boom in AI different?
The other AI winter, the funding for these science went from well funded to little funding.
I'm skeptical, with respect of course, on your statement because it doesn't have anything to back that up other than it produce useful technologies. Wouldn't this statement imply that the other previous AI which experience AI Winter (expert system, and whatever else) didn't produce useful enough technologies to have funding?
I'm currently on the camp of there is going to be an AI Winter III coming.
> None_ of them have promised wild general intelligence any time soon.
The post talk about Andrew Ng wild expectation on other things such as radiologist tweet. While it's not wild general intelligence. What I think the main article and also I am thinking is the outrageous speculation. Another one is the tesla self driving, it doesn't seem to be there yet and perhaps we're hitting the point of over promise like we did in the past and then AI winter happen because we've found the limit.
The current difference is that the technologies are actually useful right now. It's not about promised or expected technologies of tomorrow, but about what we have already researched, about known capabilities that need implementation, adoption, and lots of development work to apply it in lots and lots of particular use cases. If the core research hits a dead end tomorrow and stops producing any meaningful progress for the next 10 or 20 years, the obvious applications of neural-networks-as-we're-teaching-them-in-2018 work sufficiently well and are useful enough to deploy them in all kinds of industrial applications, and the demand is sufficient to employ every current ML practitioner and student even in absence of basic research funding, so a slump is not plausible.
Sure, keep moving timelines. It's what makes you money in the area. I am sure when around mid-2019 hits, it will suddenly be "most experts agree that the first feasible self-driving cars will arrive circa 2025".
You guys are hilarious.
Training is expensive but inference is cheap enough for Alpha Zero inspired bots to beat human professionals while running on consumer hardware. DeepMind could have released thousands of pro-level games if they wanted to and others have: http://zero.sjeng.org/
I am 100% in agreement with the author on the thesis: deep learning is overhyped and people project too much.
But the content of the post is in itself not enough to advocate for this position. It is guilty of the same sins: projection and following social noises.
The point about increasing compute power however, I found rather strong. New advances came at a high compute cost. Although it could be said that research often advances like that: new methods are found and then made efficient and (more) economical.
A much stronger rebuttal of the hype would have been based on the technical limitations of deep learning.
I'm not even sure how you'd go about doing that. You could use information theory to debunk some of the more ludicrous claims, especially ones that involve creating "missing" information.
One of the things that disappoints me somewhat with the field, which I've arguably only scratched the surface of, is just how much of it is driven by headline results which fail to develop understanding. A lot of the theory seems to be retrofitted to explain the relatively narrow result improvement and seems only to develop the art of technical bullshitting.
There are obvious exceptions to this and they tend to be the papers that do advance the field. With a relatively shallow resnet it's possible to achieve 99.7% on MNIST and 93% on CIFAR10 on a last-gen mid-range GPU with almost no understanding of what is actually happening.
There's also low-hanging fruit that seems to have been left on the tree. Take OpenAI's paper on parametrization of weights, so that you have a normalized direction vector and a scalar. This makes intuitive sense for anybody familiar with high-dimensional spaces since nearly all of the volume of a hypersphere lies around the surface. That this works in practice is great news, but leaves many questions unanswered.
I'm not even sure how many practitioners are thinking in high dimensional spaces or aware of their properties. It feels like we get to the universal approximation theorem and just accept that as evidence that they'll work well anywhere and then just follow whatever the currently recognised state of the art model is and adapt that to our purposes.
Who's to say we won't improve this though? Right now, nets add a bunch of numbers and apply arbitrarily-picked limiting functions and arbitrarily-picked structures. Is it impossible that we find a way to train that is orders of magnitude more effective?
Currently, people are projecting and saying that we are going to see huge AI advances soon. On which basis are these claims made? Showing fundamental limitations of deep learning is showing we have no idea how to get there. How to get there yet, indeed, just we have no idea how to do time travel yet.
The end result of this advancement to our world is earth shattering.
On the high compute cost. There is an aspect of that being true but we have also seen advancement in silicon to support. We look at WaveNet using 16k cycles through a DNN and offering at scale and competitive price kind of proves the point.
Current AIs have limitations but, at the tasks they are suited for, they can equal or exceed humans with years of experience. Computing power is not the key limit since it will be made cheaper over time. More importantly, new advances are still being made regularly by DeepMind, OpenAI, and other teams.
Unsupervised Predictive Memory in a Goal-Directed Agent
Moravec's paradox is the discovery by artificial intelligence and robotics researchers that, contrary to traditional assumptions, high-level reasoning requires very little computation, but low-level sensorimotor skills require enormous computational resources.
What do you think of recent papers and demos by teams from Google Brain, OpenAI, and Pieter Abbeel's group on using simulations to help train physical robots? Recent advances are quite an improvement over those from the past.
Now using models for RL is the obvious choice, since trying to teach a robot a basic behavior with RL is just absurdly impractical. But the problem here, is that when somebody build that model (a 3d simulations) they put in a bunch of stuff they think is relevant to represent the reality. And that is the same trap as labeling a dataset. We only put in the stuff which is symbolically relevant to us, omitting a bunch of low level things we never even perceive.
This is a longer subject, and a HN is not enough to cover it, but there is also something about the complexity. Reality is not just more complicated than simulation, it is complex with all the consequences of that. Every attempt to put a human filtered input between AI and the world will inherently loose that complexity and ultimately the AI will not be able to immunize itself to it.
This is not an easy subject and if you read my entire blog you may get the gist of it, but I have not yet succeeded in verbalizing it concisely to my satisfaction.
Why not petaflops of bytes then?
You mean Monte Carlo Tree Search, which is not at all like Ma(r)kov chains. You're probably mixing it up with Markov decision processes though.
Before criticising something it's a good idea to have a solid understanding of it.
The biggest minds everywhere are working on AI solutions, and there's also a lot in medical/science going on to map brains and if we can merge neuroscience with computer science we might have more luck with AI in the future...
So we could have a draught for a year or two, but there will be more research, and more breakthroughs. This won't be like the AI winters of the past where it lay dormant for 10+ years, I don't think.
The first generation TPUs used 65536 very simple cores.
In the end you have so many transistors you can fit and there are options on how to arrange and use.
You might support very complex instructions and data types and then four cores. Or you might only support 8 bit ints, very, very simple instructions and use 65536 cores.
In the end what matters is the joules to get something done.
We can clearly see that we have big improvements by using new processor architectures.
“The new spring in artificial intelligence is the most significant development in computing in my lifetime.”
He listed many examples below the quote.
“understand images in Google Photos;
enable Waymo cars to recognize and distinguish objects safely;
significantly improve sound and camera quality in our hardware;
understand and produce speech for Google Home;
translate over 100 languages in Google Translate;
caption over a billion videos in 10 languages on YouTube;
improve the efficiency of our data centers;
help doctors diagnose diseases, such as diabetic retinopathy;
discover new planetary systems;
An example from another continent:
“To build the database, the hospital said it spent nearly two years to study more than 100,000 of its digital medical records spanning 12 years. The hospital also trained the AI tool using data from over 300 million medical records (link in Chinese) dating back to the 1990s from other hospitals in China. The tool has an accuracy rate of over 90% for diagnoses for more than 200 diseases, it said.“
Well first off: letters to investors are among the most biased pieces of writing in existence.
Second: I'm not saying connectionism did not succeed in many areas! I'm a connectionist by heart! I love connectionism! But that being said there is disconnect between the expectations and reality. And it is huge. And it is particularly visible in autonomous driving. And it is not limited to media or CEO's, but it made its way into top researchers. And that is a dangerous sign, which historically preceded a winter event...
The difference between the current AI renaissance and the past pre-winter AI ecosystems is the level of economic gain realized by the technology.
The late 80s-early 90s AI winter, for example, resulted from the limitations of expert systems which were useful but only in niche markets and their development and maintenance costs were quite high relative to alternatives.
The current AI systems do something that alternatives, like Mechanical Turks, can only accomplish with much greater costs and may not even have the scale necessary for global massive services like Google Photos or Youtube autocaptioning.
The spread of computing infrastructure and connectivity into the hands of billions of global population is a key contributing factor.
I would argue this is well discounted by level of investment made against the future. I don't think the winter depends on the amount that somebody makes today on AI, rather on how much people are expecting to make in the future. If these don't match, there will be a winter. My take is that there is a huge bet against the future. And if DL ends up bringing just as much profit as it does today, interest will die very, very quickly.
This is analogous to the way electricity took decades to realize productivity gains in the broad economy.
That said, the hype will dial down. I am just not sure the investment will decrease soon.
So I guess we're waiting for something similar to happen with AI and then get AI 2.0?
The current road infrastructure (markings, signs) has been designed for humans. Once it has been modernized to better aid the self-driving systems, we don't probably need "perfect" AI.
AI is overhyped and overfunded at the moment, which is not unusual for a hot technology (synthetic biology; dotcoms). Those things go in cycles, but the down cycles are seldom all out winters. During the slowdowns best technologies still get funding (less lavish, but enough to work on) and one-hit wonders die, both of which is good in the long run. My friends working in biology are doing mostly fine even though there are no longer "this is the century of synthetic biology" posters at every airport and in every toilet.
Those are actual features that are available today to anyone, that were made possible by AI. Do you think it would be possible to type "pictures of me at the beach with my dog" without AI in such as short time frame? Or to have cars that drive themselves without a driver? These are concrete benefits of machine learning, I don't understand how that's biased.
If there are 100 facts that indicate a coming AI winter, and Brin just talks up the 15 facts that indicate AI's unalloyed dominance, that's definitely biased.
Second, I'm not quite sure that's how it works. Like in mathematics, if your lemma is X, you can give a 100 examples of X being true, but I only need a single counter-example to break it.
In my opinion a single valid modern use-case of AI is enough to show that we're not in an AI winter. By definition an AI winter means that nothing substantial is coming out of AI for a long period of time, yet Brin listed that Google alone has had a dozen in the past few years.
You cannot ask a generic question, then attack the answer based on absence of evidence for a specific example.
>translate over 100 languages in Google Translate;
>caption over a billion videos in 10 languages on YouTube;
barely even work. Yeah, it's a difficult problem but it's not even close to being solved.
Google's captioning works well when people speak clearly and in English. Google translate works well when you translate well written straightforward text into English. It's impressive but it's got a long way to go to reach human grade transcription and translation.
I think when evaluating these things people underestimate how long the tail of these problems is. It's always those pesky diminishing returns. I think it's true for many AI problems today, for instance it looks like current self-driving car tech manages to handle, say, 95% of situations just fine. Thing is, in order to be actually usable you want something that critical to reach something like 99.999% success rate and bridging these last few percent might prove very difficult, maybe even impossible with current tech.
Every time I activate it I am in for a good laugh more than anything actually useful.
Have now, ah! Philosophy,
Law and medicine,
And unfortunately also theology
Thoroughly studied, with great effort.
Here I am, I poor gate!
And I'm as smart as before
It's unable to 'understand' that 'I poor gate' makes no sense at all.
Google Translate is the 'poor gate'.
You may think that you can now read German news, but in fact you would not know if the sentence meaning has been preserved in the English translation. The words itself might look as if the sentence makes sense - but the meaning is actually shifted - slight differences, but also possibly the complete opposite.
The translation also does not give you any indication where this might be and where the translation is based on weak training material or where there is some inference needed for a successful translation.
Maybe true but they are words that are about things which are either true or not true. Has nothing to do where the words were shared. Saying they are on an investment letter so not relevant seems very short sighted.
But just looking at the last 12 months it is folly to say we are moving to a AI winter. Things are just flying.
Look at self driving cars without safety drivers or look at something like Google Duplex but there are so many other examples.
Using the list provided, one example
"caption over a billion videos in 10 languages on YouTube;" - This doesn't say how accurate the captions acutally are. In my experience youtube captioning even of english dialect isn't exactly great. For one example try turning on the captions on this https://www.youtube.com/watch?v=bQJrBSXSs6o
so it's true I'm sure to say they've captioned the videos AI based techniques, but that doesn't mean they're a perfected option.
Also (purely anecodtally) Google translate also isn't exactly perfect yet either...
This is notorious with current technology: you can demonstrate anything. A few years ago Tesla demonstrated a driverless car. And what? Nothing. Absolutely nothing.
I'm willing to believe stuff I can test myself at home. If it works there, it likely actually works (though possibly needs more testing). But demo booths and youtube - never.
The BICEP2 fiasco is a good example why.
Edit: also, a blog post with more examples and a link to the related publication: https://ai.googleblog.com/2018/04/looking-to-listen-audio-vi...
The original google report was discussed here a few weeks ago:
This is one of the areas I’m most enthusiastic about but … it’s still nowhere near the performance of untrained humans. Google has poured tons of resources into Photos and yet if I type “cat” into the search box I have to scroll past multiple pages of results to find the first picture which isn’t of my dog.
That raises an interesting question: Google has no way to report failures. Does anyone know why they aren’t collecting that training data?
I’ve assumed that the reason is the same as why none of the voice assistants has an error reporting UI or even acknowledgement of low confidence levels: the marketing image is “the future is now” and this would detract from it.
what is this 'understand'?
For most things, that people dream of and do marketing about need another leap forward, which we haven’t seen yet (it’ll come for sure)
Also, while a lot of these can be seen as "improvements", in many cases, that improvement put it past the threshold of actually being usable or useful. Self-driving cars for example need to be at least a certain level before they can be deployed, and we would've never reached that without machine learning.
Utterly useless. And I don't think it is improving.
As for the users, sure the translation may not be perfect, but I'm sure if you were deaf had no other way of watching a video, you would be just fine with the current quality of the transcription.
At least in English, they are now good enough that I can read without listening to the audio and understand almost everything said. (There are still a few mistakes here and there but they often don’t matter.)
I tried to help a couple channels to subtitle and the starting point was just sooo far from the finished product. I would guess I left 10% intact of the auto-translation. Maybe it would have been 5% five years ago; when things are this bad 100% improvement is hard to notice.
It is super cool how easy it is to edit and improve the subtitles for any channel that allows it.
If by 'almost everything', you mean stuff that a non native English speaker could have understood anyway, then yes.
The vast majority of English learners are not able to caption most Youtube videos as well as the current AI can.
You underestimate the amount of time required to learn another language and the expertise of a native speaker. (Have you tried learning another language to the level you can watch TV in it?)
Almost all native speakers are basically grandmasters of their mother tongue. The training time for a 15-year-old native speaker could be approx. 10 hours * 365 days * 15 years = 54,750 hours, more than the time many professional painists spent on practice.
A weak speaker may use a cognate, idiom borrowed from their native tongue or a similar wrong word more often. The translation app produces completely illegible word salad instead.
Deep learning is the method of choice for a number of concrete problems in vision, nlp, and some related disciplines. This is a great success story and worthy of attention. Another AI winter will just make it harder to secure funding for something that may well be a good solution to some problems.
Nobody thought to ask: "How do you know all of that content is terrorist content? Does anyone check every video afterwards to ensure that all the blocked content was indeed terrorist content?" (assuming they even have an exact definition for it).
They might not, but they could sample them to be statistically confident?
We log everything and are even starting to automate decisions. Statistics, machine learning, and econometrics are booming fields. To talk about two topics dear to my heart, we're getting way better at modeling uncertainty (bayesianism is cool now, and resampling-esque procedures aged really well with a few decades of cheaper compute) and we're better at not only talking about what causes what (causal inference), but what causes what when (heterogeneous treatment effect estimation, e.g. giving you aspirin right now does something different from giving me aspirin now). We're learning to learn those things super efficiently (contextual bandits and active learning). The current data science boom goes far far far far beyond deep learning, and most of the field is doing great. Maybe those bits will even get better faster if deep learning stops hogging the glory. More likely, we'll learn to combine these things in cool ways (as is happening now).
I'd contend for the general public, AI is a synonym for machines like: HAL; The Terminator; Star Trek's "Data"; the robots in the film "AI"; and so on.
We're nowhere remotely in the vicinity of that, and no-one even has any plausible ideas about how to start.
A random person outside of tech probably doesn't even know what deep learning is. They might have heard of it somewhere in passing.
AI is a superset and Machine learning is a subset of AI and most funding is in deep learning. Once Deep Learning hit the limit I believe there will be an AI winter.
Maybe there will be hype around statistic (cross fingers) which will lead to Bayesian and such.
Ok, with that out of my system, no, Bayesian methods are definitely not a subset of deep learning, in any way. Hierarchical Bayes could be labeled "deep Bayesian methods" if we're marketing jerks, but Bayesian methods mostly do not involve neural networks with >3 hidden layers. It's just a different paradigm of statistics.
He sees the latent layer in the hierarchical model as the hidden layer and the Bayesian just have a strict restrictions/assumptions to the network where as the deep learning is more dumb and less assuming. A few of my professor thinks that PGM, probability graphical model is a super set of deep learning/neural network.
This is where my thinking come from.
IIRC, a paper have shown that gradient descent seems to exhibit MCMCs (blog with paper link inside that led to this conclusion of mine: http://www.inference.vc/everything-that-works-works-because-...).
But I am not an expert in Neural Network nor know the topic well enough to say such a thing. Other than was deferring to opinions of some one that's better than myself. So I'll keep this in mind and hopefully one day have the time to do more research into this topic.
Things like the German tank problem or the problem of hardening airplanes during WW2 have that very AI'esque feel to it. Where you use data to build a model, then let that data from the model to change the model as it fits.
Also the whole thing about 'decision making' is either bayesian or frequency based models in nature. Most of these algorithms and math has long existed before the current boom.
Its just that the raw computing power and resources that you have today make it possible for you to deal with large amounts of data to stress test your models.