Foto van een onderzoeker

onderzoek

Laudatio uitgesproken door prof. dr. Jan Cornelis

LAUDATIO Prof. dr. PATTIE MAES
Proximus: Prof. dr. Jan Cornelis

Software Agents, Human-Computer Interaction, User Modelling, Artificial Intelligence, Artificial Life, Simulation of Adaptive Behaviour, Community-Enabling Technology – and more specialized concepts like memory-based learning, evolution-based programming, multi-agent collaboration, collaborative filtering, and negotiation algorithms… these are all keywords describing the research activities of Professor Pattie Maes. Rodney Brooks (Director of the Artificial Intelligence Laboratory, specialist in Humanoid robotics and Artificial life), Marvin Minsky (brilliant computer scientist and mathematician – Author of the book “The Society of Mind” in which he presents his conception of human intellectual structure and function), John McCarthy (key member of the original Al intelligentsia), Nicholas Negroponte (famous digital futurist and director/founder of the MIT Media Lab, author of the 1995 best-seller, “Being Digital”, which has been translated into more than 40 languages) – these are just a few representative names of the colleagues within the research environment of Pattie Maes at the MIT Media Laboratory.

  • 1987: Doctor in Computer Science at VUB, “Computational Reflection” in the area of Al and human-computer interaction,
  • 1987-1991: Senior Research Scientist of the Belgian National Science Foundation, associated with the Al Laboratory of the VUB (directed by Prof. Luc Steels), coordinator of the project “ ACognitive Architecture for an Autonomous Agent”,
  • 1989-1991: Visiting Professor at MIT’s Artificial Intelligence Laboratory, and later a joint Research Scientist appointment at MIT’s Media Laboratory and Artificial Intelligence Laboratory,
  • Fall of 1991: With a tenure-track faculty position at the MIT Media Laboratory, she becomes Assistant Professor,
  • 1995: Associate Professor, MIT, Media Laboratory.

These are some of the major milestones in the academic career of Dr. Maes. Some examples of spin-off companies illustrating the activities of Pattie Maes as an active Internet entrepreneur since 1994 are the following: Firefly Network, Inc., one of the first companies to commercialise personalization and profiling technology acquired by Microsoft in 1998; Open Ratings, Inc., a provider of performance data on businesses for improving the effectiveness of B2B e-commerce, using advanced data bases that dig deeply into individual transactions on Web sites; Frictionless Commerce, Inc., another B2B e-commerce company that grew out of her research group; Lexeme, Inc., which commercialises concept-based navigation technology. Dear Professor Maes, ladies and gentleman, many people we meet in our professional activities are outstanding researchers, Doctors, Ph. D’s. The degree of Doctor Honoris Causa, however, is a very special distinction that is reserved to persons transcending the barriers of their own scientific discipline, exploring and mastering diverse skills and knowledge domains, and bringing down virtual but sometimes tenacious barriers between scientific disciplines.

We – the VUB – have already recognized the excellent research capabilities of Pattie Maes in the domain of Artificial Intelligence by awarding her the PhD title. Today we take pride and pleasure in conferring on her the Honorary Doctorate, for the impact of her ideas and scientific work on society and the real world. Navigating between the theoretical and the practical, Dr. Maes has demonstrated that Al’s promise is more than an academic exercise. She is the prototype of the New Economy scholar/entrepreneur: a university scientist committed to seeing her ideas becoming socio-economic successes. Pattie Maes has engineered scientific breakthroughs in software agents, an original technology producing major changes in human-computer interaction and Internet usage.

Research... In almost 20 years as a researcher, Pattie Maes has charted new research fields at least three times. Each time she became recognized as one of the leaders in the field. The first such research area was that of Computational reflection, which was the topic she investigated in her PhD thesis at VUB in the Artificial Intelligence Laboratory ARTI, coordinated by our colleague Luc Steels. When she started this work, there were just a few independent individuals doing work on meta-architectures and reflection. She sensed that this was an important new topic in Programming Languages. The thesis listed the important issues in this field and contributed the first object-oriented language with reflective capabilities.

The second research area that she embarked on was that of Behaviour-Based Artificial Intelligence. Again, at the time at which she became interested, there were just a few researchers pursuing this radical and (at the time) controversial approach to AI. She contributed to the field with original algorithms for action selection and learning. This field is still very active and growing.

The third and most recent field that she helped create is that of Software Agents. She organized the first symposium on software agents at MIT in 1992 and attracted an audience of 1100 people. Her research group built many of the first prototypes of software agents and developed new techniques that have now been widely adopted.

She was invited by Scientific American to write an article on software agents for their 150th anniversary issue, a special issue on “the technologies of the 21th Century”. She was invited as one of 16 visionary speakers at the conference celebrating the 50th anniversary of the ACM (Association for Computing Machinery). Dr. Maes quotes her Software Agents work as her most important research achievement.

She is indeed widely regarded as one of the pioneers of this research and applications area. Her celebrity was cemented in the last few years when both Time and Newsweek dubbed Dr. Maes a member of the “cyber elite” and a name to watch in the new millennium. The encapsulating discipline of her work on Software Agents is referred to as “autonomous agents” research. The term “agent” denotes a fairly abstract concept. Generally speaking, an agent is a computational system that inhabits a complex, dynamic environment. The agent can sense, and act on, its environment and has a set of goals or motivations that it tries to achieve through its actions. Depending on the type of environment, an agent can take different forms. Autonomous robots are agents that inhabit the physical world; computer-animated characters inhabit simulated 3-D worlds; software agents inhabit the world of computer events and networks. The same basic questions are studied in these different domains: How does an agent decide what to do so as to progress toward its goals; how does an agent learn from experience; how can multiple agents collaborate? Pattie Maes has developed new techniques and algorithms that address these issues and she has built prototype systems that demonstrate the applications of this technology. She has worked on all three forms of agents and their applications. In particular, she worked with Rodney Brooks on developing architectures for mobile robots, with applications such as surveillance, exploration and mining.

The Brooks approach was more pragmatic than the prevailing paradigm of recreating cognitive processes. He built small autonomous systems to perform simple tasks. Through personalisation, mining and filtering their databases these systems adapt to conditions by repeating successful decisions, thus yielding practical, if not intelligent results.

In her early years at the Media Laboratory, Pattie Maes also started the “ALIVE” project together with Bruce Blumberg, Trevor Darrell and Alex Pentland, with focus on building agents, i.e. “believable characters” that live in computer graphics environments. Soon after she joined the Media Laboratory she realized that agents, which operate in digital environments, offer more interesting and challenging applications than their hardware counterparts. A software agent is a computational system which inhabits the environment of computers and computer networks and which assists users with their daily computer-based tasks.

Visionaries like Nicholas Negroponte and Alan Kay have long argued that agents and delegation should be crucial concepts in the interface of tomorrow, but few results had been achieved in this area when Pattie Maes started working on it in the early nineties.

In 1991, Pattie Maes founded the Software Agents Group at the Media Laboratory. The mission of the group was to build software agents that people delegate tasks to. The motivation for this work was that, as people’s lives get more complex and specifically as we use the computer to access more information and deal with more and more tasks, the current dominant metaphor of human-computer interaction, i.e. direct manipulation, would not prove sufficient.

Software agents differ from current-day software in that they are proactive (taking the initiative to help the user by making suggestions and/or automating the more mundane tasks the user normally would have to perform), adaptive (learning the user’s preferences, habits and interests as they change over time), personalized (customizing their assistance according to what they have learned about the user), and autonomous (operating with minimal supervision on behalf of the user).

Pattie Maes proposed a novel approach to building software agents which relies on statistics and machine learning techniques, rather than adopting the knowledge-based approach which a few other researchers had pursued earlier, with limited success. The software agent learns how to assist a user by “watching over the shoulder” of the user and noticing certain patterns and regularities in the user’s habits or interests that can then be automated. In addition, since many users have similar habits and interests agents can learn how better to assist their users by exchanging information with other agents that assist like-minded people. What makes this approach to the problem particularly appealing is that (1) it results in personalized software agents, (2) it minimizes efforts on the part of the user, (3) it results in a continuously adaptive and improving solution and (4) it allows the user and the agent gradually to build up a relationship of trust.

Pattie Maes’ research team has built several software agents for a range of applications:

  • Eager assistant agents, which find patterns in the user’s behaviour and then offer to automate them.
  • Information filtering agents, which learn a person’s interests and suggest relevant new information.
  • Remembrance agents, which log everything a user does and then later help the user remember past actions, people and situations.
  • Matchmaking agents, which help introduce people to others who have similar interests.
  • Agents that buy and sell on behalf of users, actively negotiating deals to transact goods and services.

The developments of these prototypes has yielded advances in the state of the art in AI techniques such as memory-based learning, evolution-based programming, multi-agent collaboration, collaborative filtering, and negotiation algorithms. Research summary...

A professional Statement of P. Maes, December 1997: “My research interests started out in Artificial Intelligence. Up to about five years ago, my main goal was to understand intelligence by building computational systems that demonstrated “intelligent behaviour”. After I joined the Media Laboratory I became increasingly interested in researching combined forms of human and machine that demonstrate “superior intelligence”, rather than building standalone intelligent systems. My work focused increasingly on intelligence augmentation (IA) rather than AI. Indeed, the goal of my software agent research was to expand peoples’ capabilities by allowing them to delegate to software entities. For example, we focused on software that can help a person remember things, software that can act as “extra eyes and ears” that look out for a person’s interests, software that can help a person cope with large amounts of information, and so on”. Spin-off companies...

Long before the Internet was a commercial medium, Pattie Maes saw the opportunities offered by combining her ideas about intelligence augmentation and the Web technology.

In the mid-1990’s, the commercial capabilities of software agents were clear to Dr. Maes and her students. The work of Pattie Maes attracted attention outside academia due its impact on e-commerce models.

The first commercial release of software agents occurred in 1995, when Pattie Maes launched Firefly Network. The collaborative filtering technology implemented in software agents offered the ability to provide an individual customer with customized recommendations for products based on his or her past purchasing or browsing behaviour.

Now almost all good portals and retail sites use some form of collaborative filtering. The software agents of a following spin-off company Frictionless Commerce are designed to do one thing more, namely help buyers decide how to buy.

A new focus of attention for Pattie Maes is business-to-business (B2B) e-commerce. Her latest venture Open Ratings offers ratings services for e-commerce sites. Although Open Ratings is not an agent in itself, it is a crucial link for software agents that are actually empowered to do buying and selling on a company’s behalf. Some current research and goals for the next few years...

Community Ware – “Science has long ago found solutions for physical and sensory shortcomings – glasses, hearing aids, pacemakers, artificial limbs. Why not develop prostheses for the mind. There are just so many things that people are not good at, for example multitasking, keeping track of many things at once, doing extensive searches and remembering things” says Pattie Maes. “A lot of my work has been about developing and using sets of fairly simple software agents to augment people’s cognitive capabilities. The software agent is a dynamic entity that can be acting on our behalf while we are talking or sleeping or working.” With an announced next project, known as Community Ware, Pattie Maes intends to exploit the combination of two facts: the fact that people are increasingly networked wherever they are, and the fact that very few of the problems we solve as an individual are truly original (in probably 99.9% of the cases, whatever problem a person is currently dealing with has been dealt with before by someone else).

In this new research phase she shifts from augmented intelligence to a field that she describes as follows: “Basically what I would like to focus on for the next several years is what technology and infrastructure we need, to allow people to learn from each other and rely on each other in such a way that we all benefit and feel “super intelligent” because we can “expand” our brain to include the knowledge, experience and skills of others. Specifically, I propose to build an electronic market for knowledge, and to have agents that can match up students and experts and make deals for an on-the-fly assistance relationship.
To participate in this knowledge marketplace, a person can create selling agents that know what expertise that person has and when that person is available to help someone else (and possibly how much s/he charges). Someone in need of help can create a buying agent that knows what kind of expertise that person needs and by when (in the next three minutes, in the next three days, etc). The buying agent may also have information on how to compare different people offering the expertise using criteria such as: price charged, reputation of the expert, level of expertise and availability. The market would automatically match up these different buying and selling agents.

The electronic knowledge marketplace blurs the distinction between teachers and students. A person can be an expert/teacher in some areas and a student in others. A typical person would have several buying as well as selling agents at any moment in time. Market forces will push the system towards equilibrium where the time and efforts of a true expert are optimally used for just those questions that cannot easily be answered by anyone else.

Technical issues that I will address are: ontologies for describing expertise, protocols for agent negotiation, reputation mechanisms, incentive mechanisms for collaboration and knowledge sharing, and efficient marketplaces enabled by technology.” The knowledge marketplace is just one example of what Pattie Maes terms community ware or technology that allows communities of people to organise themselves, to be more effective and collaborate in totally new ways. It illustrates an emerging trend brought about by computer networks, namely the increasing power shift from current authorities that control our lives to decentralized communities organizing themselves.

“One of the reasons why I want to research community ware is that I believe in the positive impact of such technology on our society. For example, the knowledge marketplace application described above will value currently undervalued people in society such as the elderly (they in fact will have the most experience and knowledge to sell, and the most available time).

I am hoping to emphasize the benefits of my research on community ware by applying it in some of the major problems areas of our society: health, education, poverty, and so on. As in my previous research, I have a strong desire to make my work in this area be directly meaningful to society and to have it make a large impact.

Ultimately, what excites a researcher is not only to get approval from your peers, but on top of that to get your stuff out there and have impact.”

Congratulations... Professor Maes, I have tried to summarize both your scientific work and your success in making this directly meaningful to society. You are working in one of the areas for which the second half of the 21st century will probably be cited in the history books – the first signs of artificial intelligence and thinking machines or agents showing some properties that might be qualified as semi-intelligent behaviour.

I hope that in this short time, I have highlighted the main characteristics of your work and professional career:

  • The visionary aspect, i.e. your talent to spot the right scientific subject at the right time,
  • The real impact on society,
  • Your perseverance and exemplary role in bridging the gap between scientific theory and practice.

For these reasons, the VUB takes pride and pleasure in conferring on you its Honorary Doctorate.