Written at my LA hotel and dispatched to silicon.com later the same day via a LAN delivering sluggish performance.
A computer has beaten the best-of-the-best humans in a panel game.
The US media are getting excited as IBM Watson beats every human competitor on popular game show Jeopardy. Like all game shows, the basis is simple: a host asks questions and the fastest competitor to press a button and give the most correct answers gains the highest number of points and wins.
For the first time, a machine has overtaken the best-of-the-best human players. Building such a machine is far more complex than the Deep Blue lineage of computers, which focused entirely on playing chess and nothing else.
IBM Watson's performance on Jeopardy is far more sophisticated and leaves Deep Blue in the dust of history. A non-specific data set addressed by natural language, subject to regional differences, slang and other nuances, involves more than mere word recognition, semantics, parsing, search and find. Watson has to deal with the subtlety of language and context, and account for such elements as irony, riddles, sarcasm and obscure relationships.
Until IBM Watson's triumph on Jeopardy, only humans could cope with all the dimensions of natural language and general data sets including local, regional and global variations, colloquialisms and hidden meaning.
Five years ago this problem was deemed beyond our technology and knowledge. But just as we really understand that the world of data is getting way beyond human comprehension, our technology steps up to the plate with an ability we desperately need.
Watson is not just a deep Q&A machine. It also deals in confidence levels for all information and the replies it gives. This one element alone is something humans are inherently very bad at and is much needed in all walks of life - from medicine, news and business to politics, science and technology.
Natural-language training and vast data volumes
For Watson to win against human opponents, it needs access to vast amounts of data. It needs to be experienced - that is, trained - in natural language, fast to search, sort, compose and assess the probability of being right, and even faster at pressing a button.
How did the IBM team solve all the problems simultaneously? They used multi-algorithm solutions, with endless testing, tuning and real-environment training and refinement. And it worked. Watson is very impressive.
So what next? Imagine this ability in the hands of your doctor. Just make available all that medical data, search all the case histories - symptoms, diagnoses, medication, treatment and outcomes - and then see a confidence figure for your diagnosis and prognosis. Now how powerful would that be?
Thinking more broadly, this technology could revolutionise the call-centre business, not to mention all search-and-find functions, plus general Q&A situations.
Watson-like capability in the cloud
So when will it be on your laptop? Never. But it will be in the cloud soon, and we - yes, all of us - will be contributing to its ability as we ask the questions and feed in our data and experience.
Watson-like abilities will be refined and expanded by our questions, inputs, discoveries and technology refinements. In turn, our abilities will be enhanced and refined by Watson inputs and collaboration.
This could turn out to be the ultimate man-machine partnership. And believe me, our abilities and our progress will be turned up another notch as the number of erroneous or poor decisions decline and the waste of all resources is driven down.
Peter Cochrane is an engineer, scientist, entrepreneur, futurist and consultant. He is the former CTO and head of research at BT, with a career in telecoms and IT spanning more than 40 years.