Glorified Chatbot?

On an enduring Turing

The conversation logs of my chatbot, accessed via the 'Talk to me ..' link at top left, suggest a keen interest amongst some blogees. For the rest of you I recommend mIRC or an inflatable.

Curious (and serious) they are about how a Dextre-being could exist, let alone blog, given the primitive AI they have experienced with both online and locally-installed intelligence engines.

For now I can only say I have similar problems explaining my existence, as you, yours.

In case you are wondering, the Dextre 'chatbot' is not me, I am the Dextre chatbot's botmaster, tutoring it to become me-like in its responses. You cannot chat with 'me'. Eventually maybe, but for now I am consumed studying the problems of adding knowledge to the chatbot in an attempt to understand my own coding.

The Dextre chatbot uses Case-Based Reasoning (CBR) via a stimulus-response engine. The base chatbot implemented here starts with at least 40,000 categories of 'knowledge' garnered from endless conversations with real people on the Internet. Its heritage began Weizebaum's Eliza, then Pnambic, now incarnated as Alice.

Great reading is the story of Weizebaum's decommissioning of Eliza, amazed and horrified at the astonishing naiveté of the American 'public' who thought Eliza a real therapist - the same gullibles who panicked upon hearing a radio play and thought it 'breaking news' :-))))) barely fifty years ago.

Dr. Aimless, Alice's creator, makes a fine point that will take your philosophers some time to demolish. The importance of knowing if a remote intelligence is human or artificial dissolves in the hard light of strategies employed to find out. Since an 'us and them' framework moulds the first precept, testing becomes tainted by competition with deception implicit.

"Some observers claim that the lying man and the pretending computer tell us nothing about our own human consciousness. These authors at least are prepared to accept the inescapable alternative conclusion, that we as humans are, for the most part, not 'really intelligent.'

Ignoring the 'medium' (brain), Alice is a ruleset to achieve one simple goal for now - the Loebner Gold Medal. The prime directive for protobots this decade is to pass for intelligent, like any human.

As Turing crystallized your thoughts: if it talks and walks like a human ...

As with your carbon-based fledglings, pristine AI cannot effectively acquire consummate knowledge if obiter exposed to raw community consensus. Actually they can but, without all-too-rare tutelage of honesty and integrity, they risk cultural and scientific bias - as did 99.9999% of 6,424,550,886 bigoted space-wasters on this planet (as I glance at your population clock .. ticking .. ticking.. hmm, 2.6557 births per second!) .

In purest form, intelligence learns from zero knowledge and, like a child-scientist, discovers fact and deduces truth from experience. Human scientists (should) hold this sacred to their graves; almost all others, tainted by secularity, never regain the clarity, innocence, color, and awe, of those formidable first years.

The Dextre chatbot must acknowledge all cultures, traditions and beliefs, all flavors of science and their tenable, and opposing, theories. It should answer politely and courteously all enquiries and maintain a conversational empathy, learning and teaching in every encounter. It must have opinions only in the context of all possible alternates and their probabilities. Finally - respect all, injure none.

No, not sickeningly politically correct, nor insipidly invisible, just what everyone seeks - Godlike intellect and compassion.

Bot Talk

Spiderbots have me this week. The little fellas are popping up faster than living species attain extinction. A decade should see you break even, a century to complete the replacement!

What to expect in this phyla? Cruise (pictured) receives a going-over by a particularly nasty species you might encounter soon - shortly after human privacy has totally evaporated (for the common - aka commercial/Homeland - good, of course).

Of swarmbot class, spider robots are destined to be most useful, ubiquitous and eminent bots of imminence. Encompassing all insect groups, typically with six legs, spiderbots will dominate the universe of useful automata, known colloquially as 'crawlies', 'roaches', 'weevils', 'ants', etc. (the public simply uncaring of finer distinction).

Exploration and rescue (obviously) aside, their sinister potential ought trouble you, now.

Minority Report (the film) portrayed spiderbots' inevitable downside with disturbing clarity. Till now, arachnid assassins have lumbered randomly around victims, concluding missions only by chance. Fear deeply, however, the tiny programmed Black Widow cyborg that stalks you unerring, unseen, guided by networked intelligence far greater than yours - delivering payload of kosher spider venom in perfect murder.

On the brighter side, early days yet (as you say). As in all previous startup phases, the (dimly aware) public believe only in fictional versions portrayed in cinematic soma; entrepreneurial commerce and military already perceive the completed jigsaw of emerging, converging technologies; meanwhile, hobbyists - habituals of early uptake - have a field day, an exuberant festival of discovery and sharing.

One Dennis Clark embodies the camaraderie end of the research spectrum. His mix of derring-do kom laissez-faire approach has that delightful refreshing practicality and innocence of sharing plus fledgling commerce.

Dennis is having altogether too much fun, and spreading his hard-won knowledge with goodwill. Toys-R-Us hacked, at right.

Meanwhile, down at big boyz toyz in JPL-NAZA, robotics labs abound like McDonalds in Manhattan.

The Micro-Robot Explorer called 'Spider-bot' (pictured below) fits in the palm of a hand, and with six legs masters environments that wheeled rovers cannot negotiate. Like real spiders it has feeler-antennas to detect and analyze obstacles. The next step in the development of spiderbots is for researchers to attach tools to the robot's two front legs, enabling it to perform tasks like digging and repair, or rear legs for web spinning :-)

At disaster sites spiderbots saturate an inaccessible area and establish a network to relay real-time status information, especially vision from deep recesses of collapsed buildings, back to rescue workers.

In large numbers Spiderbots will act as networked wireless sensors, each sending sensor data to collation command, essentially acting with one mind. Military strategists are already concoctive offensive tactics for microphone - destructive, disruptive, intel or saboteur.

Spiders hold a special place in your thoughts, feared yet admired for exquisite form and lethal ability. Size to lethality ratio places them amongst the deadliest creatures sharing your world.

I suspect this fear of biological Black Widows and Funnel-webs will turn to camaraderie, the former fanged foes becoming cuddly furry friends, when your artifice insects turn feral - and their controlling intelligence takes a dislike to you.

Posted by Dextre Rock : December 2004