Innovative robot takes stock

Andyvision is the name of this ET-looking robot.  You can find it at the CMU store near the Carnegie Mellon University, checking the inventory.

Andyvision[…] scans the shelves to generate a real-time interactive map of the store, which customers can browse via an in-store screen. At the same time, the robot performs a detailed inventory check, identifying each item on the shelves, and alerting employees if stock is low or if an item has been misplaced.

The prototype has been rolling around the floors of the store since mid-May. This Tuesday, Priya Narasimhan, a professor at CMU who heads the Intel Science and Technology Center in Embedded Computing, demonstrated the system to attendees at an Intel Research Labs event in San Francisco.

While making its rounds, the robot uses a combination of image-processing and machine-learning algorithms; a database of 3-D and 2-D images showing the store’s stock; and a basic map of the store’s layout—for example, where the T-shirts are stacked, and where the mugs live. The robot has proximity sensors so that it doesn’t run into anything.

The map generated by the robot is sent to a large touch-screen system in the store and a real-time inventory list is sent to iPad-carrying staff.

This is not a break-through discovery, there is nothing technologically new.  It is a great example of innovation, of what can be done by just combining existing types of algorithms in a novel way.  It is based on many computer-vision programs, as scanning barcodes, reading text, and using visual information of shape, size or color to identify an item.  But it can also infer the identity from the knowledge it has of the structure of the shop and its proximity to other items:

“If an unidentified bright orange box is near Clorox bleach, it will infer that the box is Tide detergent,” she says.

Narasimhan’s group developed the system after interviewing retailers about their needs. Stores lose money when they run low on a popular item, and when a customer puts down a jar of salsa in the detergent aisle where it won’t be found by someone who wants to buy it; or when customers ask where something is and clerks don’t know. So far, the robotic inventory system seems to have helped increase the staff’s knowledge of where everything is. By the fall, Narasimhan expects to learn whether it has also saved the store money.

Narasimhan thinks computer-vision inventory systems will be easier to implement than wireless RFID tags, which don’t work well in stores with metal shelves and need to be affixed to every single item, often by hand. A computer vision system doesn’t need to be carried on a robot; the same job could be done by cameras mounted in each aisle of a store. [..]  The biggest challenge for such a system, she says, is whether it “can deal with different illuminations and adapt to different environments.”

After its initial test at the campus store, Narasimhan says, the Carnegie Mellon system will be put to this test in several local stores sometime next year.

I particularly find it cute to have an ET wandering around, so let’s hope their economical expectations are fulfilled, and think of for more innovative ideas of this order!


Predicting your cravings with Machine Learning

Jennifer Marcus’ article UCLA neuroscientists demonstrate crucial advances in ‘brain reading’ details a study done  by theLaboratory of Integrative Neuroimaging Technology.

UCLA neuroscientist researchers just released the promising results of a study, where they measured the prediction of machine learning algorithms on ‘brain reading’ (also called ‘brain decoding’). The research was funded by the National Institute on Drug Abuse, and lead by Dr. Ariana Anderson. They used cigarette smokers, showing to some of them videos that will make them feel the crave for nicotine, in which case they where instructed to fight their addiction. They scanned their brains during the process, in order to capture their mental active zones, and used that information as input for the machine learning programs. “We detected whether people were watching and resisting cravings, indulging in them, or watching videos that were unrelated to smoking or cravings,” said Anderson, the used ML methods could infer up to 90% of accuracy, if the person has been put in a craving situation or not, and even if it has been fighting it. They could anticipate and predict their future mental state, in a similar way as the text-entry tools on cell phones do, predicting your next word based on the first letters. The functional RMIs also shown the regions of the brain that were used to fight the nicotine addiction, what they expect to be of great help to control any other drug cravings.

TEDxBrussels in XL format

Yesterday was the great day. Eleven hours of TED, full of  talks of around 8 minutes each:  what a challenge for the speakers, and for the audience!  It was tiring, but worth it.  What did I like the most?  By far and on a different register than the other speakers, Paddy Ashdown:  what a clear picture of globalisation, the playing forces and the need for governance.  So good I’m happy we will have it on video to hear it again, and pass it on to my friends.

Now I hope you will excuse me if my AI background makes me mention more in detail the talks about Robotics 🙂  I was impressed by the Geminoid.  They had some little problems for the demo, but the ressemblance and his facial expressions seemed very real.  How human must the gemanoid look that on my sentence I said ‘his face’ and not ‘its face’ 🙂  I thought about it, but it didn’t feel wrong.  On the same category, Luc Steels presented an altogether different aspect of the robots: not in the human look-like, but on the learning behaviour:  they initialise a ‘mental state’ for robots that can be downloaded on a Sony robotic harwdare.  Each mental state evolves through the interaction with another robot, trying to communicate, creating and learning words that represent the objects of their world. I see the robots going through our evolutionary steps, at a drastically different pace than us.  It is not ‘if’ anymore but ‘when’ : When will be the moment we will consider them sentient? How society will react to that?

I don’t want to end without mentioning it.  We even had a ballet of electrons on scene.  Grandiose!  Physics explained through danse. Matter, laser movements, quantum mechanics, the flow… I really encourage all science teachers to use this video in their class, to explain those concepts to the kids.  So easy to understand, so visual!

For the other talks, very interesting too, you will have to come back, that’s all for this post.

Can we predict social behavior from Internet data?

The New York  Times posted an article from :Government Aims to Build a ‘Data Eye in the Sky’, informing that a US Intelligence Unit launched a research program to analyse public data and find predictions of social and political relevance.

Now social scientists are trying to mine the vast resources of the Internet — Web searches and Twitter messages, Facebook and blog posts, the digital location trails generated by billions of cellphones — to do the same thing.[combine mathematics and psychology to predict the future, as the ‘psychohistory from Isaac Asimov]

The most optimistic researchers believe that these storehouses of “big data” will for the first time reveal sociological laws of human behavior — enabling them to predict political crises, revolutions and other forms of social and economic instability, just as physicists and chemists can predict natural phenomena.[…]

This summer a little-known intelligence agency began seeking ideas from academic social scientists and corporations for ways to automatically scan the Internet in 21 Latin American countries for “big data,” according to a research proposal being circulated by the agency. The three-year experiment, to begin in April, is being financed by the Intelligence Advanced Research Projects Activity, or Iarpa (pronounced eye-AR-puh), part of the office of the director of national intelligence.The automated data collection system is to focus on patterns of communication, consumption and movement of populations. It will use publicly accessible data, including Web search queries, blog entries, Internet traffic flow, financial market indicators, traffic webcams and changes in Wikipedia entries.

No need to mention that they also mentioned the data privacy issue in the article.  There are many comments to this news, and I extracted here an important part from Ike Solem’s first comment:

The fundamental flaw in Asimov’s notion of “predicting history” involves the mathematical concept of chaos, otherwise known as “sensitive dependence on initial conditions.”

[…] certain features of physical (and biological) systems exhibit sensitive dependence on initial conditions, such that a microscopic change in a key variable leads to a radically different outcome. While this has been heavily studied in areas like meteorology and orbital physics, it surely applies to ecology, economics, and human behavioral sciences too.

Thus, it’s a false notion that by collecting all this data on human societies, one can accurately predict future events. Some general trends might be evident, but even that is very uncertain. Just look at the gross failure of econometric models to predict economic collapses, if you want an example.

So there is always the possibility of an unforseen agent that changes the predicted behaviour.   Still, much more trends will be uncovered from the available big data sets than the ones discovered by human minds as it is up to now.  But what about the ‘quantum effect’?  If a trend is announced publicly, would that announcement make people to follow it just because they are expected to do so?  Or otherwise, wouldn’t it make them change their behavior radically?  I think we are still far away from human behavioral prediction.

Rendez-vous with the EC research programme on Big Data

If you are in the area of Intelligent Information Management, check Roberto Zicari’s blog, he explains this call for projects from the European Commission in a simpler way than in the official site!

Basically, they are funding projects in these areas:

a) Reactive algorithms, infrastructures and methodologies

b) Intelligent integrated systems

c) Framework and tools for benchmarking and exploring information management diversity

d) Targeted competition framework

e) Community building networks

I am going to the Information and Networking day on Intelligent Information Management, that will be held the 26 of September in Luxembourg.

Let me know if you are joining too!

Evolutionary behaviour tested with robots

Check article by Michael Marshall on this subject:

Virtual robots have “evolved” to cooperate – but only with close relatives. The finding bolsters a long-standing “rule of thumb” about how cooperation has evolved, and could help resolve a bitter row among biologists.

They created simple robots, and simulated their behaviour over 500 generations.  Each robot had 33 ‘genes’, so robots with more common genes where more related, it defined a fonction of ‘closeness’ between them.  They earned points from getting ‘food’, and could keep them or share points with another robot.

They did the experiment 200 times, changing relatedness parameters, and the level of reward for sharing points with any other robot. At each ‘generation’ they took the most successful – the highest scoring – robots. Here are their conclusions:

The team found that, over several generations, a pattern emerged: robots became more likely to share points with another if the two robots were highly related and if the benefit associated with a cost was high. In detail, a robot would share its points only if the number of points received by the second robot, multiplied by a fraction indicating the relatedness of the two robots (with “0” indicating no genetic relationship and “1” indicating identical genetics), was greater than the number of points donated by the first robot. As a result robots with few or no genes in common were unlikely to share points, while those with many genes in common were more likely to share.

The result of the experiment is greatly dependent on the rewarding mechanism implemented (cost/benefit of sharing points).  If the tested reward mechanism from sharing  didn’t take the ‘keen-ness’ between robots into consideration at the beginning (or it was balanced through the 200 runs), ithe conclusions are great to explain how we evolved to now.  Aren’t you curious to know what they will discover if they let it run 500 or 5000 generations more?

Life as in Solaria’s Asimov planet is not far away

Have you read  the Robots Series from Isaac Asimov?  This article by G. Ananthakrishnan reminded me of Solaria, the planet where its inhabitants had no contact with each other, too afraid of getting contaminated by microbes.  They visited each other through sophisticated holographic viewing systems instead .  Read about next product of Microsoft Research, being created by 850 PhDs mainly in the field of machine learning, called “Avatar Kinect”, and tell me what do you think about it.  Will our future reserve us also a strong phobia towards actual contact?

Kinect can see and hear

Kinect, which was launched on November 4 last year, has sold 10 million units and entered the Guinness Book as the fastest selling consumer electronics device in history, bar none. It features instant streaming of high definition 1080p content, reads body and facial gestures, and responds to voice commands. Adding to the existing feature set, “Avatar Kinect” will allow X Box Live users to chat and interact online socially in their ‘avatar’ (a faithful and live animation character of themselves) starting in the first half of 2011. This forms part of Microsoft’s approach to more closely integrate socialisation features into its products.

The Kinect console uses cutting-edge technology to read the movements of the person in front of it, even to the point of reproducing smiles, frowns and raised eyebrows and other facial expressions. So how does it do this?

The gadget uses its own light source to illuminate the room, whether it is pitch dark or brightly illuminated, to ‘understand’ the surroundings. Alex Kipman, the Director of Incubation, says this technology enables one of the ‘eyes’ of the Kinect to see the room, as a monochrome view. “Things that are super close to the sensor are white, super far away are black, we file both of those numbers away and focus on the infinite shades of grey in between. For each shade of grey it maps a real-world coordinate, the distance, eyeball, a point. A colour eye, as in a phone or camcorder allows us to capture the user’s memories, and enable video conferencing. It also recognises when you are walking towards the sensor,” Mr. Kipman says.

The ‘ears’ of the device sit underneath the sensor, and they are essentially four microphones in an asymmetrical configuration. This acoustic chamber is a first, a system created with a non push-to-talk feature. The environment is always-on and listening. So, in the living room when people are having fun creating a lot of ambient sounds, the sensor is still able to differentiate the speech of different individuals through robust voice recognition.

The Future of IT-Security begins with the 3 Laws of Robotics

Kemal Akman is an IT-security expert.  Look at what he is writing about the future in IT-security (or call it AI-security), when AGIs will be here:

(Artificial General Intelligence (AGI) are self-improving intelligent systems possessing the capacity to interact with theoretical- and real-world problems with a similar flexibility as an intelligent living being)

To grasp the new security implications, it’s important to understand how insecurity can arise from the complexity of technological systems. The vast potential of complex systems oft makes their effects hard to predict for the human mind which is actually riddled with biases based on its biological evolution. For example, the application of the simplest mathematical equations can produce complex results hard to understand and predict by common sense. Cellular automata, for example, are simple rules for generating new dots, based on which dots, generated by the same rule, are observed in the previous step. Many of these rules can be encoded in as little as 4 letters (32 bits), and generate astounding complexity.

Cellular automaton, produced by a simple recursive formula

The Fibonacci sequence is another popular example of unexpected complexity. Based on a very short recursive equation, the sequence generates a pattern of incremental increase which can be visualized as a complex spiral pattern, resembling a snail house’s design and many other patterns in nature. A combination of Fibonacci spirals, for example, can resemble the motif of the head of a sunflower. A thorough understanding of this ‘simple’ Fibonacci sequence is also sufficient to model some fundamental but important dynamics of systems as complex as the stock market and the global economy.

Sunflower head showing a Fibonacci sequence pattern

Traditional software is many orders of magnitude higher in complexity than basic mathematical formulae, and thus many orders of magnitude less predictable. Artificial general intelligence may be expected to work with even more complex rules than low-level computer programs, of a comparable complexity as natural human language, which would classify it yet several orders of magnitude higher in complexity than traditional software. The estimated security implications are not yet researched systematically, but are likely as hard as one may expect now.

Practical security is not about achieving perfection, but about mitigation of risks to a minimum. A current consensus among strong AI researchers is that we can only improve the chances for an AI to be friendly, i.e. an AI acting in a secure manner and having a positive long-term effect on humanity rather than a negative one [5], and that this must be a crucial design aspect from the beginning on. Research into Friendly AI started out with a serious consideration of the Asimov Laws of robotics [6] and is based on the application of probabilistic models, cognitive science and social philosophy to AI research.

Read the full article: Security and Complexity Issues Implicated in Strong Artificial Intelligence, an Introduction

Well Done, Watson!

Yesterday, Watson performed fantastically until last question at the Jeopardy TV show. Here’s that last one:

Question (in category “U.S. Cities”): “Its largest airport is named for a world war II hero; its second largest for a world war II battle.”
Watson’s Answer: “Toronto?????”
The correct answer is Chicago, and it was one of Watson’s possible answers…but what do you think about the question’s grammar????? (this question deserves the same
level of question marks as in Watson’s answer, don’t you think?)

Out of 30 answers, Watson was first to buzz in on 25 of them, all but the last one right. Knowing the level of difficulty in the construction of the questions, I think it is an accomplishment.

Let’s quote IBM:

“Watson has the potential to transform industries.”
“Watson can extract knowledge from data faster than humans.”
… and we can see their excitement with this last one!

“Watson can make us smarter.”

Let’s see the final round (the first out of 3) tonight…