Empathetic Circuits

A Cambridge team’s computers sense emotions

By Morgon Mae Schultz

As computers become more efficient and powerful, technology permeates into areas of life that may seem unlikely beneficiaries of small, fast processors—like human emotion. At the University of Cambridge computer laboratory, a team of researchers is addressing what it calls the necessity that computers become socially and emotionally intelligent, developing computers that can sense and react to user’s emotions. Possible applications range from improved automobile dashboards to aids for those unable to interpret emotional cues due to autism. The Cambridge lab’s graphics and interaction team, led by Professor Peter Robinson, has tested systems that can infer a user’s emotional state from facial expressions, gestures and voice as accurately as the top 6 percent of humans.

Reading Minds

The ability to discern what another person is feeling, known in psychology as mind reading, crosses cultural boundaries. Scientists agree on the facial expressions that reveal six basic human emotions (happy, sad, angry, afraid, disgusted and surprised) as well as hundreds of subtler mental states (such as tiredness, joy or uncertainty). Robinson’s team uses probabilistic machine learning to train computers that recognizing visual cues such as head tilt, mouth pucker and eyebrow raise. Such a system could inform a car when its driver is bothered, upset, bored or drowsy.

The team has applied the same mind-reading capabilities to an “emotional hearing aid,” a portable device designed to translate facial expressions into emotions and suggest appropriate reactions so people on the autism spectrum can relate to those around them. MIT is pursuing emotional-social prostheses, and Robinson says his team continues to research interventions for autism-spectrum conditions.

The inference system is as accurate as most people—70 percent. “There is potential for improvement. In some sense computers already are better than individual people, but there will always be difficulties establishing the ground truth in a subjective assessment,” Robinson says.

On a larger physical scale but more intimately customized, a gesture-reading system lets users control music through emotional body postures, creating an interactive, real-time soundtrack. Because emotional expression in gestures varies widely among individuals, it’s harder for machines to read whole-body cues than facial expressions. The system must tune itself to each new user and, just like humans, it reads large, dramatic movements more easily than subtle everyday gestures.

A Robot Named Charles

Not content with computers being able to recognize our mental states, Robinson’s team is working on machines that can synthesize emotion—express feelings in a way that triggers humans’ natural understanding. The complexities of human-human interaction present huge challenges to designing an appropriate human-robot interaction. One pioneer is Charles, an animated android head that Hanson Robotics built for the Cambridge team. Aiming for more satisfying robot experience, Charles has a plastic face (modeled after its mathematician namesake Charles Babbage) that can smile, grimace, raise its eyebrows and otherwise express a range of emotions.

Robinson says his team is always seeking commercial channels that could bring its technologies to consumers, including a major car manufacturer that may implement the emotional-inference system. “We are always talking to companies about the possibilities for commercial exploitation. I guess that something will appear when they see a good business case.”

You can see people interacting with Cambridge mind-reading machines here.

This video contains footage of Charles.

---

Morgon Mae Schultz is a copy editor for MSP TechMedia.

Align IT With Business

By Natalie Boike

The upcoming SHARE conference in Anaheim, Calif., (Feb. 27–March 4) is aimed at helping IT workers better align their departments with business needs, all with an eye towards continuing costs. “The thought behind the topic is to ensure we’re not losing sight of the relationship between IT and business,” explains SHARE President Janet Sun. “Over time IT has gotten a bad wrap as being inwardly focused and using new technology without focus on the business. I don’t think that’s always true.”

Related subtopics include IBM’s the zEnterprise System, virtualization, application technologies and architectures, and security and privacy. Each subtopic will relate back to the main theme. For example, Sun says virtualization sessions will outline not only how the technology can make better use of existing resources, but also how it can be a cost-effective approach to delivering new services.

“A lot of our sessions focus on user experiences and draw out best practices, from real-world practitioners,” Sun says. “We believe our attendees will take information from SHARE and it will help them contain costs and better align with business objectives.”

Keynote Sessions

As usual, SHARE has lined up three keynote speakers who will discuss wide-ranging IT topics.

Anjul Bhambhri, IBM VP of Big Data Products, will be discussing how the industry is changing to handle larger amounts of data and larger numbers of data queries. Sun says Bhambhri will discuss some of the technologies used in the IBM Watson supercomputer, which is slated to perform in a “Jeapordy!” challenge airing on Feb. 14, 15 and 16.

Dayton Semerjian, general manager at CA Technologies, will discuss the next generation of mainframe management. Additionally, IBM VP of Social Business and Collaboration Solutions and Social Media Evangelist Sandy Carter will outline how attendees can better leverage social media.

Other Highlights

New to SHARE is a conference within a conference targeted to IT executives, called ExecuForum. Held Feb. 28–March 1, discussions and roundtables are designed to deliver true best practices from executive-level peers that have applied or are about to apply strategies within their own environment. For example, topics scheduled for discussion include securing and managing mobile devices, and social media and its impact on IT, Sun explains.

SHARE will also continue its online virtual conference. Those who can’t physically travel to the event can access four days of quality streaming content and six months of archived on-demand access to recorded sessions. Sun says on-site attendees will still gain the most benefit. “Face-to-face networking and communication is one of the historic strengths of SHARE, but SHARE online is a good way to join the community,” she says.

For more information or to register, visit www.share.org.

---

Email Natalie at [email protected]

Growing Mastery

After six years, the Master the Mainframe contest is still drawing an increasing number of new users to the IBM mainframe. A record 3,537 students from more than 400 schools across the United States and Canada participated in the most recent contest; the winners of the three-part contest were announced Jan. 26.

Mike Todd, with the IBM System z Academic Initiative, says the increased participation is likely a result of the current generation of students who realize mainframes offer a solid career path. “If students didn’t see a future in mainframe computing, we wouldn’t be seeing the kind of growth that the contest has experienced: from 750 students in 2005 to 3,537 students in 2010,” he says.

Patricio Reynaga, West Texas A&M University, took first place. Another college student, Jay Thomas, Pace University, was awarded second place. A high school student, Calvin MacKenzie, Arkansas School for Mathematics, Sciences and the Arts, received third place. Other top winners and honorable mentions are listed online.

Sponsored by the IBM Academic Initiative System z program, the Master the Mainframe Contest invites students to gain hands-on experience with the System z platform. The contest welcomes students who have never logged on to a mainframe system before, guiding them through the basic tasks required to successfully navigate the system.

Three-Fold Competition

The contest is divided into three increasingly difficult parts, allowing students to decide how deep they want to go into the inner workings of the mainframe. In Part 1, students learn to navigate the user interface. In Part 2, the challenges get much more difficult; students debug Job Control Language (JCL) errors, learn to navigate UNIX on the mainframe, alter C programs to produce different output, manipulate security protocols and learn more advanced system navigation. All of the students who complete Part 2 earn certificates of completion from IBM and receive invitations to upload their resumes to the IBM Student Opportunity System (a resume database accessible to all IBM Business Partners and clients).

Todd says participants receive everything they need to log on to a mainframe system for the first time: screenshots, detailed instructions and a healthy dose of encouragement. “We understand it can be intimidating to tackle a brand new platform, and we do our best to ease students into the world of enterprise computing,” he adds. As students progress through the contest, they learn skills needed in future challenges.

In Part 3, students are faced with problems that have flummoxed systems programmers in the real world. IBM clients have also pitched in ideas for skills they’d like to see students acquire during the contest, such as more experience with VSAM data sets. “Students must bring tenacity, dedication and technical ability to succeed in the contest, Todd says, “but they don’t need to have a background in mainframe-specific technologies.” In fact, past winners have started the contest with no mainframe experience at all.

“Students see that we’re teaching them real-world skills that are in demand with many of the world’s largest companies, and they’re drawn to the contest in part because it’s fun and they can win prizes, but also because they’re picking up skills that can give them an advantage in the job market,” he adds.

Adding Rational Developer for System z

This year, for the first time, students were invited to download and install Rational Developer for System z (RDz), an Eclipse-based integrated development environment (IDE), to complete the Part 3 challenges. RDz is a more familiar environment to most students than a traditional 3270 emulator, Todd says. Contest organizers, who were predicting technical-support issues to arise, were surprised at how quickly students became productive with RDz. As no technical support-issues arose, RDz will likely be a component in future Master the Mainframe competitions.

Get Involved

The 2011 competition kicks off the Tuesday after Labor Day, with the contest beginning in early October. Students can keep up with the latest on the contest at ibm.com/university/contest, and also on the Facebook fan page.

Even if a contest isn’t running in your country (or if you’d like to practice for an upcoming one), students and educators have worldwide, no-cost mainframe access through the IBM Academic Initiative System z. Visit the website for details.

---

Email Natalie at [email protected].

Backing Up Cloud

I miss the good old days when I had maintenance windows that were long enough that I could bring my machine down to single user mode and back up the whole system. These backups contained all of the data that mattered to the company at the time. Twenty years ago, I could only back up my machine with reel-to-reel tape drives. I'd bring my machine down to single-user mode to perform the backup, and each tape backup would take 12 minutes. I remember this because we would set the time on a portable kitchen timer when we started each tape. When the timer went off, we'd head to the computer room to swap out the tape, and go to the console to “press G to continue the backup.” All of the important data lived on that one machine. We didn’ t worry about distributed computing environments, as we weren’ t running any at the time. Sure we had a few PCs scattered here and there, but they weren’ t critical. The entire company and all of its data lived on that central machine, and users who sat in front of green-screen dumb terminals accessed it. There wasn’ t any data that users stored locally; it was all stored on the machine in the computer room.

When I hear about cloud computing, this is still the kind of environment I picture: where people are logged into a central machine that exists in a computer room in the sky. I use several Web-based applications like salesforce.com, webex.com or Google Mail, where I know nothing about the servers nor where the applications run, and I don’ t necessarily care about the hardware or operating systems the applications use. I log in, use the service and log out. I often find myself logging into the IBM virtual loaner program website, where I can utilize slices of IBM hardware for short periods of time for demonstrations or proof of concepts or education.

I’ ve worked with companies that have cloud offerings, where I can very easily log in, spin up some resources on their servers and then spin them back down when I am finished with them. As long as my response time is acceptable, do I really care about the physical hardware these virtual instances run on?

I’ ve also had customers who were unable to get resources to test hardware in their environment. Using the cloud, they were able to log on to a cloud provider, spin up some server resources, do the that they needed and spin the resources back down—all without waiting for their internal IT departments to acquire and configure hardware for them. This would also benefit users who have test hardware several generations behind what they’ re using in production. Instead of using old hardware, they can use more modern machines in virtual environments as needed.

Consider This

There are benefits to cloud computing, but there may be a few things contemplate when considering a leap from your own computing assets to those that you don’ t control. I realize that these days we’ re usually accessing cloud-based applications over the Internet instead of from a green screen directly attached to computers in the machine room, but concerns like privacy, security and availability need to be considered along with all the benefits that are touted with the cloud.

Backup and recovery is another consideration when deploying services to the cloud. How do we back up our data that lives in the cloud? Surely cloud providers offer snapshots and local backups, and maybe that’ s good enough for what you’ re doing. If you wanted to copy your data to machines that are under your control, would you use the network and some kind of continuous data protection in order to move data from the cloud to machines you own so that you have another copy of it? Or would that method of data protection defeat the purpose of having someone else handling infrastructure management?

What happens if somewhere down the road you decide you want to get out of the cloud? Are there going to be issues with getting your data or OS images back under your control? Can you easily clone the systems back onto your own hardware or will you be looking at server reloads?

I have watched customers struggle with liberating their data from outsourcing companies and contracts. The companies that manage the machines have custom tools and scripts that they don’ t want to hand over. They may have information around how the machines were configured that they don’ t want to share. What’ s your plan to get out of the cloud or move to another cloud provider if you find the one you are using isn’ t for you? What do you do if the service you’ re using goes down, or the company goes out of business or they change the interface so much that you no longer like the way you use the tool? Will upgrades and outages happen on your timetable or on theirs? When you get used to accessing servers and applications from anywhere there’ s a network connection and then you find the provider has an outage, you want to be sure providers offer information and status updates on when they expect to recover the systems.

I enjoyed reading a blog post from John Scalzi, who was trying an experiment where he would exclusively use Google Docs and a Google laptop computer to write a novel. Technical glitches began causing delays, and he eventually retuned to working from his desktop, saying, “Until ‘the cloud’—and the services that run on them—can get out of your way and just do things like resident programs and applications can, it and they are going to continue to be second-place solutions for seriously getting work done.”

Return to Centralization

While there are definite advantages to the cloud-computing approach in some situations, I can’ t help but think that the whole idea has a “Back to The Future” feel to it, where we take distributed computing resources and try to centralize them again, or worse yet, rebrand existing offerings as cloud offerings so we can say we’ re on the cloud bandwagon. Certainly there are going to be applications and situations that will benefit from moving applications out of data centers. We just need to be sure to do our homework and educate ourselves before making the leap.

---

Rob McNelly is a System p Solutions Architect for Meridian IT Inc. and is a former administrator for IBM. Rob can be reached at [email protected].