July of 2000
ICMAS - International Conference of Multi-Agent Systems
Friday the 7th to 13th
We came to Boston to attend possibly the last big ICMAS (International Conference of Multi-Agent Systems) Conference - I submitted a paper on the Digital Self, but it didn't get up, alas ... but I handed out a few copies to interested people I meet - and the ATAL Workshop before they get amalgamated (along with Autonomous Agents - AA) into something even grander (2002: Yes, together the three of them now form AAMAS, with the first joint annual event in Bologne, Italy - with Toru Ishida as the General Chair). Its six days of agent research, agent technology and agent languages. Most of the agent luminaries are here including but not limited to: Barbara Grosz from Harvard ('Discourse Structure and Planning: Collaborative work, team coordination, collaboration and competition'), Mike Wooldridge, Victor Lesser, Phil Cohen ('Center for Human Computer Communication' / Social commitment vs Internal commitment), Carles Sierra, Milande Tambe, Frank Dignum (Agent Theory for Team Formation), Michael Luck, Mark d'Inverno, Liz Sonenberg (BDI meets Deontic events), Sarit Kraus, Katia Sycara, Munindar Singh, Toru Ishida, and many others.
The conference venue is The Boston Park Plaza, where we are staying - so no excuses for late arrival to morning sessions. And the Cheers Bar (the set for the comedy that saved the reputation of US comedy in the outer world, before Seinfeld) is just across the park. There's some twilight happening, the light stays well into the evening, so C and I head out each evening after the conference sessions, to explore Boston and surrounds.
Tom Malone Keynote:
Tom Malone was one of the keynotes, and his talk was impressively broad-ranging. "People are Agents, but Agents aren't people" says Tom. "And therefore agents should always report to a person who is responsible for their actions". Also, this raises the issue of communication between agents and people to a heighened level of importance. He cited the CHI conference with its attendance at ~2500 and CSCW with ~ 500 as examples of the interest in the area of human-computer interaction. Even so, "there is room for fundamental theories in the coordination of managing dependencies amongst activities" involving humans and agents - "flow, sharing and fit" of activities and resources - "right time, right place, right thing". He mentioned an MIT Process Handbook with some 5000 processes documented.
Speaking of which, we checked out the MIT campus across the Charles River in neighbouring Cambridge (apparently it used to be in Boston, known as Boston Tech, until it moved across the river in 1916). See photo 1. On the way we passed some sort of music bowl in the park on the southern bank of the river. Jazz with a touch of the operatic. Made a mandatory visit to the MIT Press Bookstore in Main Street, to checkout their full range of titles. Came across Leon's 'The Art of Prolog' book - which gave me my first real appreciation for the quality MIT Press books can reach. Also saw the bland yellow covered Constraint Programming by Stuckey and Marriot on the shelves, which I also used as a text as a student of logic and constraint logic programming. Boston seems to be blessed with a lot of independent bookstores, although one of the more famous ones - Waterstones Booksellers, built in local 'brownstone' (common in the Back Bay area), see photo 2 - closed its doors last year, which says something, but I not sure what. We also checked out some of the Harvard campus while in Cambridge. See photo 3. They have an interesting early computer called the Harvard Mark 1 (it appears that IBM built it, but Harvard rebadged it?) housed in one of their less grandiose buildings (Cabot Science Building)... but not as good as the one we have in Melbourne J , it could play music (i.e. called CSIRAC it was functionally completed in 1951 in Sydney, lived at the Uni of Melbourne 1956 to 1964, and is destined for the Melbourne Museum 'real soon now'. Postscript 2001: Installed in the Melbourne Museum on public display, this year).
|Photo2:Waterstones Booksellers in the Back Bay area||Photo 3 : One of the better looking buildings on Harvard campus|
Back to Tom Malones keynote ...He talked of the "decreasing cost of communication" is fast making it feasible to supply "all information to all points" in time and space. He mentioned the 'E-Lance Economy' (electronic freelance), where small (1 to 10 person) firms will be the norm, often formed dynamically for the project at hand. Examples already abound in the film and construction industries.
Tom says that "the complexity of our computational systems will be roughly equal to a mouse's brain by 2009 (I hope he means the fury ones), a human brain by 2019, and by 2099 more than the whole of humanity! And no one is really in control" ... what's he trying to do, empty this agent conference out? (On the other hand, that's probably a safe bet, as the intellectual wisdom of humanity on the whole, probably balances out fairly low e.g. WW2 certainly didn't end war before the 21 st century. The green revolution didn't end poverty. Higher education rates don't seem to improve voting habits...) And he left us with several questions including: "How can we do better?" "How much better can we do?" ... how long is a piece of string? "While AI has suffered from exaggerated expectation, the AI winter is over."
Back across the river in Boston proper there is the 'newer' part of town to the north of the park (Back Bay), built on reclaimed swamp (Boston apparently used to have two other hills). Its come a long way - see photo 4.
|Photo 4 : Typical street in the Back Bay area||Photo 5 : The old and new often side-by-side|
And further still, are some modern office towers amongst some older buildings (see photo 5) and a great shopping precinct (check out Newbury Street, if you're in town). Right off to the north west of the shopping precinct there's some unusual modern buildings and a very long reflection pond (in which tall people can take a good long look at themselves) which is apparently the headquarters of the Scientologists - not highly advertised on the tourist information we had at hand.
There's been some good presentations here at ICMAS. Craig Boutiliar's paper on 'Decision Theoretic Planning'. He uses Markov Decision probablilities, to replace large state spaces ('state' and 'action' spaces) with much fewer equations ("only 28 parameters versus a 256 transition matrix"). Reminds me of the individual choice modelling econometrics back in my transport research days. Its always too easy to fall for a simple equation or two, that invariably misses some important assumptions. You really have to calibrate your models well, in econometrics.
There's been some interesting one-liners on the role of 'roles' in agent systems, eg:
In a 'Projects with JADE paper: "A role model is a template for agent acquaintance and interaction." - I'd agree with the interaction part, but have reservations about the acquaintance part.
In one of the poster papers: "Role is like an agent type" ... yes, that's the expression that used to irritate David Kinny, regarding agent role.
Frank Dignum had some interesting things to say:
He makes use of "conversation theories" within team formation. A travel agent enlists several local agents to achieve a task. "Contract-net agents assume that there are a lot of agents for a bid. Some times there is only one agent." Types of dialog: persuasion, negotiation, information seeking, enquiry, deliberation, evistics (when things get out-of-hand). He is doing: Beliefs (KD45), Goals (K), Intentions (KD), Collective Beliefs, Collective Intentions. His agent architecture has: reasoning module, planning module, communication module, social reasoning. His Goal: to divide modalities over separate modules with little controlled interaction. "Potential recognition: the Initiator believes there is a group of agents able, willing and have the opportunity to form a team to achieve a particular task. Assertions lead to Beliefs. Communication leads to Commitment. Declaratives lead to Facts." In conclusion: "Dialogues can be used in team formations. The moves in dialogue can be represented in Speech Acts. However, using dialogue puts a high load/claim on agent reasoning."
Liz Sonenberg has some interesting points regarding BDI meets deontic events: Norms (standardized behaviour) and Obligations (contracts between agents). The selection mechanism is "loose". In answer to several questions from the audience: She "uses PDL - a logic of preferences. Goals approximate Norms. Obligations and Norms are indexed. Commitments and internal obligations are close, but not the same."
We checked out downtown Boston and the harbourside. Boston is one of the oldest and most historically significant cities in the US. On the weekend the tall ships arrived from New York (which they sailed out of on the 4 th of July) - these conference organisers really know their stuff J There were an estimated 1 million people around the Boston harbour - that's more than the modern population of Boston itself (about 580,000 - smaller than I'd have guessed before coming here), so there was a lot of shoulder rubbing, but everyone we came across were very friendly, including the intoxicated. We bought some tall-ship teeshirts and an Irish tin whistle for the kids.
The Old State House is down here near the harbour, surrounded by tall modern buildings. See photo 6:
Built in 1713 (and renovated a few times), it was from the east balcony of this building that the Declaration of Independence was proclaimed on July 18, 1776. Six years before, in front of another of its balconies, the Boston Massacre took place when British troupes slayed 5 unarmed locals, and three years after that, the Boston Tea Party happened out on the habour.
There's the old graveyard not far up the hill from the harbour, where most of the American revolutionaries of the time got buried, and well worth a visit. While Paul Revere is buried there, his wife Sarah's tombstone is more interesting in it style. See Photo 7 ... love that skull and crossbones motif, which is the style used on a lot of the less-prominent members of the history-making cohort.
We finished the evening with some famed seafood chowder ("served to five Presidents") - two bowls each... who needs a main course when the soup is this chunky and this good!
Mike Wooldridge Keynote.
Mike is all fired up (through-out the conference), starting with his keynote on 'Computationally Grounded Theories of Agency'. He's got a new agent book out and he is certainly a good marketer for an academic.
Cape Cod and Beyond
I've always loved the Cape Cod style of house with its attic bedroom windows jutting out from the steep shingled rooves (see Photo 8) - so we just had to go there. We hired a car - despite dire warnings from the conference organisers "not to drive in Boston during your stay, because the 'Big Dig' (that's a series of underground motorways etc, 11 years in the making, so far) caused daily disruption to the road network by changing the purpose of lanes and even changing local roads overnight!" I thought that was very strange advice for people at such a 'logically'-oriented conference, after all, to the person from out-of-town, 'all' of the roads are different, any day of the week, any year you choose. So the 'daily disruption' simply evens things up for us interlopers.
So C and I hired a car and drove down to and then through Plymouth (which seems to have more flagpoles with American flags raised, per capita, than anywhere I've ever seen - but granted, the 4 th of July wasn't many days ago), and then on to Cape Cod. C had to have her mandatory KFC, and, after enquiring about my "strange accent", they gave us a double helping of what we ordered and paid for - that'll be the last time I attempt to eat 5 large pieces of seasoned chicken. Interesting - its only when you get out of American cities and into the country towns, that you realise that most of the locals have hardly ever heard an Australian accent, as they all seem to have big trouble understanding me at all ... it wasn't a problem in Boston (nor LA (nor most of California), nor New York).
During the eating of the free portions of KFC, I suddenly felt an echo of the Thanksgiving turkey supposedly given to the Pilgrims right here in this town within cooee of this take away in fact, by the Wampanoags Indian tribe, several hundred years ago. There seems to be something of a thanksgiving backlash originating from this town - in 1970 a 'National Day of Mourning' for all American Indians, was proclaimed right here in Plymouth, some 350 years after the Pilgrims took root here. It laments the annexation of their land, despite well-meaning treaties, the massacres still occurred and the land grab was endless and merciless until the Indians were completely dispossessed. Mmm, I'm starting to worry about some of those 11 herbs and spices? On the other hand, it lessens the shame of the annilation of the aboriginals of Tasmanian back home, about an emotional millimetre, knowing that even the hollier than thou Pilgrims from Gods Own Country were doing it too (i.e. you didn't have to be of convict stock, to fully indulge in the extremes of the racist virus)... The history of this place is like the dark underbelly of the freedom and fraternity story just a little way back up the road in Boston. It reaffirms my appreciation of the great Aussie distain for Flag, Flagpoles, larger than life statues and billboards and politicians who want more of them. Not to mention extreme caution about the things done on the ground in the name of grandiose schemes in the clouds. Onwards to Cape Cod...
We stayed at the Village Inn in Sandwich (that's a town, not lunch), which has the other architecture style for which the area seems well endowed - white churches with tall pointy spires. Its hard to get a good feel for the Cape, without being invited to someone's large rural summer retreat - as they seem to take up most of the acreage, but alas, we didn't know anyone in the area, so photos with the zoom lens like photo 9, were as close as we got to these watery, green, homely paradises writ small, where kids were soaking up the summer. However, there are lots of nice publicly accessible wooden jetties and raised walkways, sometimes through salt-loving reeds - it is really a very picturesque peninsula - except for the town at the end of it, saved only by its good restaurants and street lights after dark. My God the seafood is good in this place ... why would you ever have needed a turkey?
The next day we headed south with the intention of getting to Newport, Rhode Island (to see what sort of a place leads to a Denis Connor), but instead we ventured off the main drag, to New Bedford a very old town with an interesting past I'd certainly never heard of until now. The houses are wooden, painted, many in need of a new coat of paint, but the better ones are in pretty good condition. I just had to see the 'Octagonal House' - see photo 10 . For some reason, it reminded me of the Adams Family pad - it's the greenhouse off to one side, with large exotic plants one of which looked a bit like a triffod, but we didn't watch it long enough to see if it moved.
These old houses are a small remnant of the golden age of this town see photo 11 (looks a bit FLW-ish to me)... it was, for a time, "the richest town in the world" - so the locals claim. It was the home of whaling when oil from the sperm whale "lite the lights of the world" before electricity made it obsolete. Its largely fallen into disrepair - maybe that's the bad karma from all of those dead whales? But, more likely, there is probably some strange local economic force keeping this town in an outwardly apparent ill economic state (the youth that don't get attracted to New York, probably get attracted to Boston?). Maybe its just another example of how difficult it is to step out of the shadow of former glory.
People we talk to are still having big trouble with my accent, but they understand C perfectly! That's really annoying me given that English is her second language. Although, she spent a few years studying journalism in Ohio, she has no significant American accent to her English, some people in Oz do have trouble with her accent ... but they understand her perfectly in rural US? We came across an elderly couple at a café where we settled in for lunch (Freestones Restaurant), who did understand my 'speak'. The man spent time in Australia during the second world war (basically, Australia was saved from Japanese invasion culminating in two significant confrontations: the land assault in PNG was stopped by Australian soldiers and Papuan guides on the Kakoda Trail, while the sea invasion was halted in the Battle of the Coral Sea - which was largely an American fire-power outcome ... that's the primary insight into Australian foreign policy for both major political parties, ever since), and, like quite a few US serviceman in Australia at the time, he married an Australian woman, who came back to the US with him ever more. His main memory was of Ayres Rock - "the enormous size of that red monolith was so unexpected, and its right in the middle of the Country!" - and he was most surprised to hear that it is now know by its original aboriginal name of Uluru rather than Ayres Rock.
The Whale Museum is impressive, with the skeleton of a large 100 metre Blue Whale inhouse. Always good to be reminded that these are the largest creatures ever to grace the Earth, including the dinosaurs. A full-sized blue whale has a heart the size of a small car! ...I wonder how big the largest robots will be by the end of the 21 st century - given that the birth canal doesn't represent a barrier to their generational growth rate ... but with no heart at all. That's a bad thought, all I get coming to mind is the God Soldiers in Miyazaki's Valley of the Wind. However, after Tom Malone's keynote mentioned above, I should be more worried about robot mind, rather than their physical size and the inclusion/exclusion of a heart. ...And if they ever catch the racist virus, we're all in deep trouble! (I think we should stop at the 'mouse brain' stage in 2009 ... and put it in an Apple mouse, or offer it as an upgrade to the person who came up with its single-button approach to the human-computer interface.)
Notes from the Final Agent Tools Panel.
Some good questions were raised, and points made by various panel members and interaction from the audience:
"What are useful metrics and useful tools for capturing them?"
"Why aren't current agent toolsets being used?" (not to mention architectures and methods)
"How does a tool suggest, support or enforce an agent-oriented approach?"
"Is infrastructure (the lack of) the main limitation in establishing a large agent community?"
"Are there too many 'time sinks' in the way of getting up and running in agent technology?"
"Where is the agent 'killer app'?"
"In the real world there are only problems and solutions to problems."
"We need to do things the way the rest of the world does: demonstrate added value ."
"There is still disagreement on the AO mindset - compare that to OOP. The fragmented community is good for ideas, but bad for consensus. Good ideas will always be taken up." (I'm afraid that that doesn't preclude bad ideas being taken up!)
"There needs to be a custodian of KQML."
"Agent kits are disappearing, witness General Magic and IBM's Agent Kit. Perhaps survival of the fittest suggested said someone?" No, that's not it, was the consensus.
"Anyway, toolsets need to be more robust, secure and handle exceptions better."
"Redundancy can give you robustness." (It can also lead to you dragging your bum across the floor as you walk!)
Final ICMAS Panel's Views titled: The Future of Agents.
"Agents will revolutionize middleware."
"Agents will revolutionize the use of the Web."
"Agents will revolutionize how commerce is conducted."
"Agents will radically change the way software is produced."
Mike Wooldridge countered any complacency regarding this apparent consensus regarding the assured success of the Agent Oriented paradigm, with a brief talk on - The Disappearing Agent :
It is an important research discipline, but ...
"The world is going Agent ... decentralising the focus."
"Objects are becoming Agents ... therefore there is a large 'bottom-up' community in place."
"We are too 'top-down', witness the KQML language and the FIPA community."
"We are too concerned with 'smarts' ".
"The question remains: What kind of community do we want to be?"
His parting words were: "Your role is the Guiding Spirit " ... marketer yes, evangelist, maybe.
Something tells me humanity isn't doomed for a good long while yet. Remember, HTML made the www years after SGML had been around the backblocks of the high tech towers, vying to be the markup language of choice amongst those in the know. We've got a long way to go to make the AO paradigm, a genuine, widely embraced paradigm - and it's the human-side of the agent technology equation that is yet to be taken seriously.