Machine Learning (Theory)

3/13/2016

AlphaGo is not the solution to AI

Congratulations are in order for the folks at Google Deepmind who have mastered Go.

However, some of the discussion around this seems like giddy overstatement. Wired says Machines have conquered the last games and Slashdot says We know now that we don’t need any big new breakthroughs to get to true AI. The truth is nowhere close.

For Go itself, it’s been well-known for a decade that Monte Carlo tree search (i.e. valuation by assuming randomized playout) is unusually effective in Go. Given this, it’s unclear that the AlphaGo algorithm extends to other board games where MCTS does not work so well. Maybe? It will be interesting to see.

Delving into existing computer games, the Atari results (see figure 3) are very fun but obviously unimpressive on about ¼ of the games. My hypothesis for why is that their solution does only local (epsilon-greedy style) exploration rather than global exploration so they can only learn policies addressing either very short credit assignment problems or with greedily accessible polices. Global exploration strategies are known to result in exponentially more efficient strategies in general for deterministic decision process(1993), Markov Decision Processes (1998), and for MDPs without modeling (2006).

The reason these strategies are not used is because they are based on tabular learning rather than function fitting. That’s why I shifted to Contextual Bandit research after the 2006 paper. We’ve learned quite a bit there, enough to start tackling a Contextual Deterministic Decision Process, but that solution is still far from practical. Addressing global exploration effectively is only one of the significant challenges between what is well known now and what needs to be addressed for what I would consider a real AI.

This is generally understood by people working on these techniques but seems to be getting lost in translation to public news reports. That’s dangerous because it leads to disappointment. The field will be better off without an overpromise/bust cycle so I would encourage people to keep and inform a balanced view of successes and their extent. Mastering Go is a great accomplishment, but it is quite far from everything.

Edit: Further discussion here, CACM, here, and KDNuggets.

2/24/2016

Learning to avoid not making an AI

Tags: AI,Machine Learning jl@ 2:08 pm

Building an AI is one of the most subtle things people have ever attempted with strong evidence provided by the durable nature of the problem despite attempts by many intelligent people. In comparison, putting a man on the moon was a relatively straightforward technical problem with little confusion about the nature of the solution.

Building an AI is almost surely a software problem since the outer limit for the amount of computation in the human brain is only 10^17 ops/second (10^11 neurons with 10^4 connections operating at 10^2 Hz) which is within reach of known systems.

People tend to mysticize the complexity of unknown things, so the “real” amount of computation required for a human scale AI is likely far less—perhaps even within reach of a 10^13 flop GPU.

Since building an AI is a software problem, the problem is complexity in a much stronger sense than for most problems. The effective approach for dealing with complexity is to use modularity. But which modularity? A sprawl of proposed kinds of often incompatible and obviously incomplete modularity exists. The moment when you try to decompose into smaller problems is when the difficulty of solution is confronted.

For guidance, we can consider what works and what does not. This is tricky, because the definition of AI is less than clear. I qualify AI with by degrees of intelligence in my mind—a human level AI is one which can accomplish the range of tasks which a human can. This includes learning complex things (language, reasoning, etc…) from a much more basic state.

The definition seems natural but it is not easily tested via the famous Turing Test. For example, you could imagine a Cyc-backed system passing a Turing Test. Would that be a human-level AI? I’d argue ‘no’, because the reliance on a human-crafted ontology indicates an incapability to discover and use new things effectively. There is a good science fiction story to write here where a Cyc-based system takes over civilization but then gradually falls apart as new relevant concepts simply cannot be grasped.

Instead of AI facsimiles, learning approaches seem to be the key to success. If a system learned from basic primitives how to pass the Turing Test, I would naturally consider it much closer to human-level AI.

We have seen the facsimile design vs. learn tension in approaches to AI activities play out many times with the facsimile design approach winning first, but not always last. Consider Game Playing, Driving, Vision, Speech, and Chat-bots. At this point the facsimile approach has been overwhelmed by learning in Vision and Speech while in Game Playing, Driving, and Chat-bots the situation is less clear.

I expect facsimile approaches are one of the greater sources of misplaced effort in AI and that will continue to be an issue, because it’s such a natural effort trap: Why not simply make the system do what you want it to do? Making a system that works by learning to do things seems a rather indirect route that surely takes longer and requires more effort. The answer of course is that the system which learns what might otherwise be designed can learn other things as needed, making it inherently more robust.

4/22/2015

Randomized experimentation

One good thing about doing machine learning at present is that people actually use it! The back-ends of many systems we interact with on a daily basis are driven by machine learning. In most such systems, as users interact with the system, it is natural for the system designer to wish to optimize the models under the hood over time, in a way that improves the user experience. To ground the discussion a bit, let us consider the example of an online portal, that is trying to present interesting news stories to its user. A user comes to the portal and based on whatever information the portal has on the user, it recommends one (or more) news stories. The user chooses to read the story or not and life goes on. Naturally the portal wants to better tailor the stories it displays to the users’ taste over time, which can be observed if users start to click on the displayed story more often.

A natural idea would be to use the past logs and train a machine learning model which prefers the stories that users click on and discourages the stories which are avoided by the users. This sounds like a simple classification problem, for which we might use an off-the-shelf algorithm. This is indeed done reasonably often, and the offline logs suggest that the newly trained model will result in a lot more clicks than the old one. The new model is deployed, only to find out its performance is not as good as hoped, or even poorer than what was happening before! What went wrong? The natural reaction is typically that (a) the machine learning algorithm needs to be improved, or (b) we need better features, or (c) we need more data. Alas, in most of these cases, the right answer is (d) none of the above. Let us see why this is true through a simple example.

Imagine a simple world where some of our users are from New York and others are from Seattle. Some of our news stories pertain to finance, and others pertain to technology. Let us further imagine that the probability of a click (henceforth CTR for clickthrough rate) on a news article based on city and subject has the following distribution:

City

Finance CTR

Tech CTR

New York 1 0.6
Seattle 0.4 0.79

Table1: True (unobserved) CTRs

Of course, we do not have this information ahead of time while designing the system, so our starting system recommends articles according to some heuristic rule. Imagine that we user the rule:

  • New York users get Tech stories, Seattle users get Finance stories.

Now we collect the click data according to this system for a while. As we obtain more and more data, we obtain increasingly accurate estimates of the CTR for Tech stories and NY users, as well as Finance stories and Seattle users (0.6 and 0.4 resp.). However, we have no information on the other two combinations. So if we train a machine learning algorithm to minimize the squared loss between predicted CTR on an article and observed CTR, it is likely to predict the average of observed CTRs (that is 0.5) in the other two blocks. At this point, our guess looks like:

 

City

Finance CTR

Tech CTR

New York 1 / ? / 0.5 0.6 / 0.6 / 0.6
Seattle 0.4 / 0.4 / 0.4 0.79 / ? / 0.5

Table2: True / observed / estimated CTRs

Note that this would be the case even with infinite data and an all powerful learner, so machine learning is not to be faulted in any way here. Given these estimates, we naturally realize that show finance articles to Seattle users was a mistake, and switch to Tech. But Tech is also looking pretty good in NY, and we stick with it. Our new policy is:

  • Both NY and Seattle users get Tech articles.

Running the new system for a while, we will fix the erroneous estimates for the Tech CTR on Seattle (that is, up 0.5 to 0.79). But we still have no signal that makes us prefer Finance over Tech in NY. Indeed even with infinite data, the system will be stuck with this suboptimal choice at this point, and our CTR estimates will look something like:

City

Finance CTR

Tech CTR

New York 1 / ? / 0.59 0.6 / 0.6 / 0.6
Seattle 0.4 / 0.4 / 0.4 0.79 / 0.79 / 0.79

Table3: True / observed / estimated CTRs

We can now assess the earlier claims:

  1. More data does not help: Since Observed and True CTRs match wherever we are collecting data
  2. Better learning algorithm does not help: Since Predicted and Observed CTRs coincide wherever we are collecting data
  3. Better data does help!! We should not be having the blank cell in observed column.

This seems simple enough to fix though. We should have really known better than to completely omit observations in one cell of our table. With good intentions, we decide to collect data in all cells. We choose to use the following rule:

  • Seattle users get Tech stories during day and finance stories during night
  • Similarly, NY users get Tech stories during day and finance stories during night

We are now collecting data on each cell, but we find that our estimates still lead us to a suboptimal policy. Further investigation might reveal that users are more likely to read finance stories during the day when the markets are open. So when we only display finance stories during night, we underestimate the finance CTR and end up with wrong estimates. Realizing the error of our ways, we might try to fix this again and then run into another problem and so on.

The issue we have discovered above is that of confounding variables. There is lot of wonderful work and many techniques that can be used to circumvent confounding variables in experimentation. Here, I mention the simplest one and perhaps the most versatile one of them: Randomization. The idea is that instead of recommending stories to users according to a fix deterministic rule, we allow for different articles to be presented to the user according to some distribution. This distribution does not have to be uniform. In fact, good randomization would likely focus on plausibly good articles so as to not degrade the user experience. However, as long as we add sufficient randomization, we can then obtain consistent counterfactual estimates of quantities from our experimental data. There is growing literature on how to do this well. A nice paper which covers some of these techniques and provides an empirical evaluation is http://arxiv.org/abs/1103.4601. A more involved example in the context of computational advertising at Microsoft is discussed in http://leon.bottou.org/papers/bottou-jmlr-2013.

 

 

7/10/2013

Thoughts on Artificial Intelligence

Tags: AI,Announcements,Machine Learning jl@ 12:34 pm

David McAllester starts a blog.

9/3/2011

Fall Machine Learning Events

Many Machine Learning related events are coming up this fall.

  1. September 9, abstracts for the New York Machine Learning Symposium are due. Send a 2 page pdf, if interested, and note that we:
    1. widened submissions to be from anybody rather than students.
    2. set aside a larger fraction of time for contributed submissions.
  2. September 15, there is a machine learning meetup, where I’ll be discussing terascale learning at AOL.
  3. September 16, there is a CS&Econ day at New York Academy of Sciences. This is not ML focused, but it’s easy to imagine interest.
  4. September 23 and later NIPS workshop submissions start coming due. As usual, there are too many good ones, so I won’t be able to attend all those that interest me. I do hope some workshop makers consider ICML this coming summer, as we are increasing to a 2 day format for you. Here are a few that interest me:
    1. Big Learning is about dealing with lots of data. Abstracts are due September 30.
    2. The Bayes Bandits workshop. Abstracts are due September 23.
    3. The Personalized Medicine workshop
    4. The Learning Semantics workshop. Abstracts are due September 26.
    5. The ML Relations workshop. Abstracts are due September 30.
    6. The Hierarchical Learning workshop. Challenge submissions are due October 17, and abstracts are due October 21.
    7. The Computational Tradeoffs workshop. Abstracts are due October 17.
    8. The Model Selection workshop. Abstracts are due September 24.
  5. October 16-17 is the Singularity Summit in New York. This is for the AIists and only peripherally about ML.
  6. October 16-21 is a Predictive Analytics World in New York. As machine learning goes industrial, we see industrial-style conferences rapidly developing.
  7. October 21, there is the New York ML Symposium. In addition to what’s there, Chris Wiggins is looking into setting up a session for startups and those interested in them to get to know each other, as last year.
  8. Decembr 16-17 NIPS workshops in Granada, Spain.
Older Posts »

Powered by WordPress