IBM Watson, Bob Dylan and the limits of machine intelligence
Web Analytics

IBM Watson, Bob Dylan and the limits of machine intelligence

The appearance of Bob Dylan in a YouTube ad for IBM’s Watson machine intelligence services will no doubt cause bien pensant liberals to tut again, even if he is only franking the form extending back to a long line of endorsements from Victoria’s Secret women’s underwear to both Chrysler and Cadillac cars. But this latest spot also raises some interesting questions about the nature of computer learning and its chances of capturing the wayward muse of the exceptional individual.

In the video, Watson boasts that it can read 800 million pages per second (“That’s fast,” drawls uber-cool Dylan, for it is he) but it’s what comes next that is interesting. Watson summarises the heart of Dylan’s lyrics as “time passes, love fades”, meeting another droll deadpan response of “Well, that sounds about right.”

But of course it’s not nearly enough. Not for a songwriter who doffs his cap to various religious texts, jokes, traditional songs, great works of literature and the world’s places in songs that span protest, sex, love, voyeurism, obsession, good old Tin Pan Alley and, well, pretty well everything in between in most of the idioms 20th-century music has offered.

A computer can parse a text and spot keywords, and even identify themes. It can then create an English-sounding sentence summarising those themes and explain its work. What it can’t do with any great reliability is spot the ironies in the words, sarcasm in the voice, the infinite subtleties in the grain of art, the torrent of back stories and nods hither and thither to earlier art and history. The video is funny and nicely self-referential but it also points to the very weaknesses of the computer’s scope, or at least the sheer number of dimensions that must be taken into consideration and processed when analysing such subtle work.

Machine learning has other strengths, of course. Dylan’s magpie habit of pilfering and reshaping words from books, songs and magazines is very ripe for software to identify sources. It’s possible, likely even, that in future Watson and Watson-like tools will discover whole new threads in Dylan’s world, once perceptively described by another fine songwriter, Tom Waits, as “a planet to be explored”.

Dylan is seemingly something of a technophobe: he has railed against Napster in the past and against synthetic modern music. But I can imagine him being intrigued by the new realms opened up by the new technologies, perhaps nudged by his polymath son Jesse who interviewed a Watson expert in 2012 and later directed video pieces on IBM. 

But great art is often fleeting and hard to pin down. It’s tough for even the largest computer and the smartest software to say quite what makes Dylan’s allusive lyrics and ever-changing music so attractive to the ear - and sometimes the clues can lead you to all the wrong places.

“Obviously, I'm not an IBM computer any more than I'm an ashtray,” Dylan once told Playboy magazine. “I mean it's obvious to anyone who's ever slept in the back seat of a car that I'm just not a schoolteacher.”

Good luck parsing that, Watson.

 

Also read:

How far can IBM Watson go?

Wimbledon serves tech with tennis

IBM sets Watson free to rethink the world

Bob Dylan versus globalisation

PREVIOUS ARTICLE

«Top Tips: Securing your valuable software-based intellectual property

NEXT ARTICLE

Too many spies, too few rules leave no Safe Harbor for anyone»
author_image
Martin Veitch

Martin Veitch is Editorial Consultant for IDG Connect

  • twt
  • twt

Add Your Comment

Most Recent Comments

Our Case Studies

IDG Connect delivers full creative solutions to meet all your demand generatlon needs. These cover the full scope of options, from customized content and lead delivery through to fully integrated campaigns.

images

Our Marketing Research

Our in-house analyst and editorial team create a range of insights for the global marketing community. These look at IT buying preferences, the latest soclal media trends and other zeitgeist topics.

images

Poll

Will Kotlin overtake Java as the most popular Android programming language in 2018?