Archive for February, 2011
The researchers at IBM decided to make a spectre of their latest efforts into machine intelligence through the Jeopardy game show. In essence, the game show tests the machine’s ability to understand natural language and to deliver responses that are correct. On those two tasks, it performed quite well and good enough to best human competitors of the venerable game show.
That may sound like an over generalization of the research team achievement but it is not meant to be as Watson represents the latest progress in algorithms, combined with even more powerful hardware which work together to calculate and render responses within a short period of time. In a way this is an almost natural and expected progression as search engines, which are continually crawling the web for analysis and using their immense data repositories, may very well need a more refined interface like what Watson may make possible.
At the moment internet search is largely keywords driven with little in terms of contextual meaning of what is sought and indeed better filtering of the results; it would be easy to just have your search engine present you with the first 3 responses to your query and rank them according to the likelihood of their correctness. That is certain a more general use case than what is being considered Watson’s next challenge: a diagnostic aid for doctors. This certainly makes sense as the next logical area of progression for this device; the subject area is specialized and the ability to continuously add new medical research and findings may well improve decisions that doctors make. This could perhaps even bring the medical community closer to more automated or at the very least computer aided research as recently championed by one of Google’s founders. This may not be something that Watson can do but with a few modifications and adjustments, it will be a natural extension for the device.
End of Human Intelligence Dominance?
Reading the buzz around Watson’s jeopardy victory on the internet, you would think that our dominance as the most intelligent beings is at an end however the victory, while important, does not pose any direct threat to human intelligence as yet. It is more prudent to restrain from using terms like artificial intelligence as AI has the connotation of emulating human intelligence and hence a crude approximation of the latter. Machine intelligence on the other hand can freely evolve to their own capabilities as defined by advancements in hardware and algorithms as well as programming methods.
The only time when machine intelligence can overshadow human intelligence is when the former reaches a point that it achieves sentience and as such be capable of acting as a fully independent moral agent. That in itself is a much bigger challenge than emulating human intelligence since sentience does not guarantee free will and thus rational behaviour. Before getting carried away with that train of thought, machine intelligence is far from outshining human intelligence any time soon, if ever.