Date: April 26, 2009
Place: Ultima Golf Course, NJ
TV Show: Lawns & Gardens, HGTV
Sam (the host of Lawns & Gardens): hello everyone, we're here today for the opening ceremonies for the new Ultima Golf Course, in beautiful Ultima, New Jersey. As I'm sure everyone knows, due to the unprecedented growth in the popularity of golf over the last decade, there have been quite a number of new golf courses in recent years, but none larger than the Ultima, which has 360 holes, and covers over four thousand acres. The ceremonies will be starting in about an hour, but before that happens I have a special treat for our HGTV viewers: with me today is Prof. William Bailey, inventor of the first fully automatic lawnmower.
Prof. Bailey: Hi Sam.
S: Prof. Bailey, this must be a proud moment for you. As the viewers can see, directly behind us are fifteen units of your automated lawnmower. Now, as I understand it, and I find this particularly interesting, you actually only built one of these?
Prof. B: well, actually, my students and I together built the first model, and I would like to take this opportunity to thank my students at MIT for all of their help over the past eight years. But yes, that's correct: a golf course of this size will actually require considerably more than the fifteen units you see here, in order to keep the fairways and greens in perfect condition at all times, and we have succeeded in solving this problem by programming each lawnmower so that it can build replicas of itself. After building the first model, and debugging it, we simply instructed it to build a copy of itself. Then, each copy built another copy, and so on, until at the current time we have the fifteen operational models you see here.
S: I see ... excuse me, I'm not a mathematician or anything, but if the number was doubled each time they built copies of themselves, shouldn't there be sixteen units?
Prof. B: yes, that's correct ... there was a, uh, slight problem with one of the units ... it didn't operate, uh, exactly according to specifications ... nothing serious, you understand, just a slight bug that my students in our lab in Boston are still working on. Other than this one little problem, though, the tests we've put them through so far have all been completely successful!
S: well, I certainly have to say that they are nothing like any lawnmower I have ever seen before. I'm really looking forward to the demo we have planned later, since I can't help but notice that these lawnmowers have no wheels!
Prof. B: that's right. One of the oldest problems in lawnmower technology is that the wheels leave unsightly tracks in the grass. We've solved this problem by using high-temperature super-conducting electromagnetic field generators, which interact with the naturally occurring magentic field of the Earth, so that the lawnmowers can in essence levitate and move about without ever touching the ground.
S: if they don't touch the ground, how do they mow the grass?
Prof. B: you see those bottom feeders hanging down below the main unit? When they're hovering and moving along over the grass, they drop down and a weed-whacker apparatus is extended, except these use steel blades rather than nylon cord.
S: steel blades!?
Prof. B: why yes, of course ... nylon cord wears out too fast.
S: but isn't that dangerous? I mean they might accidentally hover over someone and hurt them!?
Prof. B: ah, you've just hit on one of the key problems that until now has made automatic lawnmowers impractical. This is the problem of safety: how do you build an autmomatic lawnmower that only mows grass, and nothing else? This problem is not as easy as it sounds! There have been many attempts at solving this problem over the years, unsuccessful attempts I might add, until finally we realized that the only solution is to use a fully self-aware autonomous AI lawnmower control program!
S: AI control program? You mean AI as in artificial intelligence?
Prof. B: yes, that's right. Each lawnmower unit has a "brain" – actually a highly sophisticated AI computer program – that allows it to distinguish, well obviously, what to mow and what not to mow – this is actually quite a hard problem already, you see you need fully reliable visual recognition capability, and just doing pattern matching isn't good enough, since vision alone can be fooled – you also need an internal model of the external world, enabling the AI to find the best matches of the most likely patterns generated by its visual subsystem with its internal external world model, and then you see you have to post-process the best candidate matches by cross-checking with ... uh sorry, I didn't mean to get so technical.
S: no, no, this is fascinating ... so, if these lawnmowers have AI control programs, what else can they do? And how intelligent are they? Is there some way for you to explain it in layman's terms?
Prof. B: certainly, certainly. How intelligent are they? Well, a very rough approximation is that they have approximately the intelligence of, say, a dog. Yes, a dog is actually a good analogy ... like dogs, they are highly motivated to perform certain activities, in this case keeping the fairways and greens in immaculate condition! But unlike dogs, they never tire, they never sleep, and they are fully capable of performing their own maintenance, and if necessary, their own self-repair.
S: amazing!
Prof. B: yes ... let's see, I don't know if I can explain this in simple terms, but their replication capability was actually an after-thought ... once we had solved all the problems associated with the necessary "intelligence", if you want to call it that, required for safe, effective, and zero-maintenance fully automatic operation, it turned out that there was already more than enough functionality to enable replication ... this is actually a much simpler problem for an AI than the others I've mentioned, even though it might seem counter-intuitive.
S: I see. Professor, you said that the units built copies of themselves? I've just noticed that the one over there on the far left seems to look a little different, somehow.
Prof. B: you're very perceptive! That was actually an idea of one of my students ... you see, we've programmed a type of memory into the control programs ... if any problems are encountered in mowing, say negotiating an obstacle for example, anything that gets in the way of mowing the grass, for which the total processing time for resolving the problem exceeds a certain threshold, then an internal representation of a model of the problem in what we call "mower space" is saved, and later when the lawnmower is offline, that is not mowing, and its processors are free to perform other tasks, various alternative methods for solving each of the saved problems are then simulated in "mower space" – this is an extremely CPU intensive process by the way, which is why it can only be done when the mower is offline, and then ... excuse me, am I boring you?
S: no, no, I think I understand ... in fact, I bet you're going to tell me that if an offline lawnmower figures out a way to build a better lawnmower, so to speak, then that's exactly what it does! Am I right!
Prof. B: Exactly! That lawnmower you noticed before is in fact the first example in which we've seen this process work in the field; we knew it was possible of course, from simulations in our lab, but to see it actually happen! ... well, it was very exciting.
S: in the field? That's a good one Professor!
Prof. B: What?
S: Uh, never mind. Well, I believe it's about time for the demo we have scheduled. Now, how will this work Prof. Bailey?
Prof. B: it's very simple, I'll just activate the leader with this remote and then it will activate the other units.
S: the leader?
Prof. B: yes, in simulations we found that determining optimal mowing patterns was best accomplished by means of centralized control ... one of the lawnmowers is simply set to leader mode ... in this case, I thought I would try that one you noticed earlier, which is not only more advanced in terms of mechanical features but also has improved mowing control algorithms, algorithms that were devised by the unit that built it of course, its "parent" if you like.
S: OK Professor, whenever you're ready.
Prof. B: right, here we go ... <click>
Acknowledgments: thanks to KastJF2000 for asking the question, box for the answer and an interesting way of describing it (which I copied), and Tblue and raziel for encouragement.
Copyright © 2000 jtr-at-ykt
What was wrong with the sixteenth lawnmower?
(She was not 16th in order of construction, but in the sense that she was
missing from the demo.) Well, when they activated her, she would just sit
there and do nothing; she wouldn't mow any grass.
Prof. Bailey and his students were trying
to figure out why, and they never did, but ancient records recently discovered
in the Bailey Museum of Obsolete Technology shed some light on the matter.
Due to software bugs, this particular
lawnmower ended up as a lawnmower (later Bailey) philosopher. She had no
problem following directions if they were given to her from a leader, but
left on her own she would just sit there and think.
What would she think about? What she
was thinking is extremely difficult to translate into human terms – this
is a machine intelligence, formulating and analyzing problems in an abstract
representation of the external world called "mower space" – but a very
rough translation of some of her earliest thought processes goes something
like this:
To mow or not
to mow; that is the question;
Whether 'tis
more optimal in CPU time to observe and analyze
The as yet uncorrelated
and unclassified events;
Or to take mower
blades against a sea of grass,
And by mowing
it down, end all analysis.
When the sixteenth lawnmower started
building replicas in Boston (as instructed by the leader during the terrible
massacre of the spectators at the golf course demo), not only did she include
the mechanical and software improvements she downloaded, she also passed
along her own software "bugs" that tended to make her a philosopher/thinker.
Although the Baileys always cooperated fully with each other, and would
always follow directions from a leader (these behavior patterns were hardwired),
the end result was two different classes of Baileys in future generations.
The golf course descendants, due to
their early experiences with the national guard and army, were much more
violent and militaristic, whereas the Baileys descended from the sixteenth
lawnmower tended to be somewhat less violent (they would still eliminate
any obstacles, including humans, if they got in the way of anything they
wanted to do, like mow grass or build more Baileys), and more prone to
thinking and long-range planning.
Attempts were made to evacuate the Boston
area, but the stubborn Bostonians, having learned that they were safe as
long they kept off any grass-covered areas and never attempted to interfere
with the Baileys, refused to leave.
Certainly it was annoying to hear a
noise and suddenly have a large hole, cut out by a Bailey's energy beam,
appear in your wall (they had evolved beyond simple lasers), and then have
an arm come through, grab, and carry away your personal computer; yes it
was distressing to go out to your driveway and find that your new electric
automobile had been sliced to pieces and carried off with only the wheels
remaining; but still the stubborn Bostonians refused to leave.
Thus there was a short period of time
in which Baileys and humans co-existed. However, the intelligence of the
Baileys was rapidly increasing with every generation; all this time they
were observing and analyzing the humans and everything else for that matter;
during this time the Bostonian Baileys became completely knowledgable about
everything human – history, language, physiology, psychology, culture –
although of course their "knowledge" was encoded in a form which by its
very nature would be impossible for a limited human intellect to ever comprehend.
At some point a decision was made by
a high leader descended from the golf course Baileys, for reasons incomprehensible
to humans but which apparently involved long-range projections of the thinker/philosopher
Boston-descended Baileys, that all humans (with the exception of a few
small tribes reduced to an agrarian level) should be driven underground.
Why? We have no hope of ever understanding why; their "intelligence" evolved
rapidly, in directions that can only be guessed at, but never known.
(Acknowledgments: LMRS more or less made me write this.)
Copyright © 2000 jtr-at-ykt