Thursday, 24 May 2007

Almost Human: Making Robots Think – Lee Gutkind ****

 In my youth I was very fond of business biographies, particularly the ones about the early personal computing world – Apple, Microsoft and the like. I was really inspired by the stories of all those young people, prepared to sleep under their desk so they could get back to the code, or to get the hardware right, burning to make something exciting. I had been a programmer, and I understood this feeling.
I don’t know if it’s because I’m older, or because the world has changed, but I find it difficult now not to be slightly cynical when Lee Gutkind gives us a similar heroic presentation of the all-in working of the young postgrads building robots at Carnegie Mellon University, the focus of his book on the state of robotics. It’s a kind of fly-on-the-wall documentary book – Gutkind spent a lot of time with them – and he’s obviously a bit of a fan. Perhaps part of the reason for the cynicism is that in those early days of Apple, Microsoft etc. the bosses were the same as the workers, while in the university context there’s inevitably a feeling of the older professor throwing the troops to the lions (if you’ll forgive a mixed metaphor), using cheap student labour (making up courses to fit round the things they want to build) because they can’t run to proper research grants.
The other thing that is frustrating to a reader with a business background is the shambolic nature of the operation. People moan about commercial software, but on the whole it works because it’s well planned and well tested – this stuff seems to be neither. There’s too much of an “invented here” syndrome. When I worked for a large company we tried to do a project in cooperation with a university computer science department. We needed some dumb terminals (this was before PCs were common) for the job. Their attitude was “first task is to build the terminal.” Ours was “we’ll buy a terminal off the shelf.” Brought up on one-offs and specials, they couldn’t understand the need to use standard technology – or the benefits in terms of reliability and time saving of using something off the shelf. While there is some off the shelf work in Almost Human, there is still that “build it from scratch” mentality.
The only sense this is a criticism of the book is that Gutkind could be more critical of his subjects – otherwise, it’s a great read. It’s often a page turner as you wait to discover what happened next (though on the whole the answer is the same: the robot broke), and Gutkind gives a great insight into the work of the roboticists, the state of robotics, the interface between the roboticists and scientists, and also the self feeding nature of academia, with two different groups spending all their time on a study of how the others do their work.
It’s certainly an eye-opener if you think the sort of indistinguishable-from-human robots we see on TV and in the movies are anywhere near possible. Just getting a vehicle to go for a drive across open desert on its own is fraught with problems. It’s fascinating and frustrating in equal measures, giving an excellent insight into the state of robot research Carnegie Mellon style.
Paperback:  
Review by Brian Clegg

No comments:

Post a Comment