AI still not clever enough

PUBLISHED : Tuesday, 30 August, 2005, 12:00am
UPDATED : Tuesday, 30 August, 2005, 12:00am
 

Artificial intelligence is growing, but we're still the smartest on the planet - for now


Even today, some argue that the human mind is radically different from any other kind of intelligence - natural or artificial.


For example, Push Singh, who runs an artificial intelligence project at the Massachusetts Institute of Technology, argues that the more we look at the human mind, the more intricacy we see. At this institute, they question whether science is anywhere near concocting a 'thinking machine'.


I suspect that computers will outthink us one day, just as they already perform calculations and memorise information much better than we can.


Sure, they lack subtlety, but so what? Just look at the progress they have made in that game of strategy, intuition and cunning: chess. Way back in 1948, British mathematician Alan Turing developed a chess algorithm designed to challenge human opponents. The algorithm blew it, losing to an amateur player, and AI didn't seem like such a clever idea after all.


But it was far from a case of game over. Duels pitting man against machine became a source of entertainment and the benchmark for determining how far AI had evolved.


In the 1990s, IBM's Deep Blue went on to dazzle and defeat the grandmasters, showing that AI was advancing steadily.


Now, computer programs routinely humble grandmasters. Unless you cripple the software, you probably cannot beat your computer or PDA at chess, or for that matter, checkers.


Nonetheless, despite the hype generated by that flick in which Jude Law played silicon smoothie Gigolo Joe, AI has yet to establish itself as an integral part of our lives. The only AI that I regularly encounter scarcely merits the word 'intelligent'. The 'smart' search engine I use just seems to churn out results in brute force style, with minimal insight into what my quarry might be.


When I roam that AI wonderland, Second Life, sure the software enables me to badger, flirt with and rile other beings, human or otherwise. But it still feels worlds away from a time where AI is everywhere and so clever that it can pass the Turing test, convincingly posing as a fellow human for at least five minutes: simple conversational programs such as Eliza (www-ai.ijs.si/eliza/eliza.html) just do not cut it.


Even so, the field has by no means fallen into a stupor. Quite the reverse, according to Ray Kurzweil and Terry Grossman in their exploration of progress titled Fantastic Voyage.


Kurzweil and Grossman claim that science is advancing at a blistering pace. 'Because of this exponential growth, the 21st century will equal 20,000 years of progress at today's rate of progress,' they said. And within 25 years, AI will match our mental prowess in scope and refinement, then leave us in the dust, struggling to make sense of what happened.


Meanwhile, AI is starting to make waves where it is needed: the field of medicine.


A computer system designed by a University of Sheffield team mimics a doctor's expertise in making treatment decisions for intensive-care patients. The system will give them just the right drug fix they need, shielding them against the threat of septic shock.


Another AI medical miracle is the Rheo Knee, the sensors of which analyse the joint a thousand times a second, enabling it to adjust to any step or stumble.


The Rheo's mindfulness allows amputees equipped with it to walk much more smoothly than with a standard low-tech peg.


At New York Presbyterian Hospital in Manhattan, Penelope the nurse can 'see', 'hear' and even, through complex programming, 'reason' - in other words, make choices.


According to her maker, Michael Treat, Penelope is a colleague instead of a collection of electrical impulses: 'a stand-on-its-own kind of entity'.


In Dr Treat's view, which echoes Kurzweil's, personal robots have now reached the spot where PCs hovered 20 years ago. We just need to give AI a couple of decades before it truly deserves to be ranked as intelligent.


For now, let us indulge our lingering egomania and enjoy our status as the smartest entity on the planet while we can.


Share

 

Send to a friend

To forward this article using your default email client (e.g. Outlook), click here.

AI still not clever enough

Enter multiple addresses separated by commas(,)

For unlimited access to:

SCMP.com SCMP Tablet Edition SCMP Mobile Edition 10-year news archive