Tuesday, November 29, 2011

Do Androids Dream of Electric Sheep?

Mark Stephen Meadows is one of those really annoying Americans who has seemingly turned his hand to everything and is probably just getting started. His story would be hard to believe, had it not been so heavily documented in the form of paintings, books, lectures, websites and US patents. Meadows claims to speak English, French and (conversational) Spanish, as well as C++, Javascript and 'cocktail conversation'. In his own words, he left school to be an artist, got hooked up in early web development – Meadows did graphics design work for Stewart Brand's legendary WELL (Whole Earth 'Lectronic Link) when it became a dot com in the early 1990s – later worked at Xerox PARC and Stanford Research Institute, had his art work exhibited in Paris, and went on to found a software company called HeadCase. Something of an adventurer, Meadows traveled to Iraq in 2003 and wrote a book about his liaisons with the Tamil Tigers in Sri Lanka in 2010, Tea Time with Terrorists. He now spends his days varnishing his boat anchored off the coast of Mexico.

Like most Americans, he is not backward about coming forward (why should he be?) and there is a slightly self-mythologising quality about his remarkable career so far. But that is probably just jealousy speaking. There is no denying that the boundless enthusiasm, energy and joie de vivre apparent in such an outstanding CV is intimidating to a wannabe like me. But letting that sort of feeling overtake me is not going to help to accomplish any of the wild schemes I have in my head...

First, I will just review this book.

We, Robot is a travelogue about how technology is catching up with, and in some cases overtaking, fantasy. The book is ostensibly about robots - those humanoid worker drones invented by a Czechoslovakian playwright in the 1920s and as yet not realised - but manages to encompass a surprisingly wide range of technological phenomena. The breathless prose and sometimes stream-of-consciousness-machine-gun-of-ideas style does leave one gasping for air at times, but then you remember, in many ways, this is just an introduction, an overview - you relax and let the high winds carry you.

The book skids between the ethics of drone warfare, and remote control and autonomous weapons systems as embodied by The Terminator; the utility of Jetsons style home cleaning robots; augmentations that enhance human potential such as Luke Skywalker's hand; the odd appeal of sex-bots; and the quest for an artificial intelligence, which hopefully will not 'malfunction' like HAL in Stanley Kubrick's 2001: A Space Odyssey. I have not even mentioned Meadows musings on Battlestar Gallactica (the newer good one), the inadequacy of the Turing Test and James Cameron's Avatar - and still more ground is covered in the text itself. It is little wonder therefore that the wanderer does occasionally lose his way, and, although much is learned during those diversions, the book is strongest when addressing the central idea that motivated his writing in the first place.

In an online interview here Meadows says that he wrote the book to 'explore some ethical questions' - and it is in those moments that We, Robot really soars. The book provides an excellent description of the Uncanny Valley, a concept with which anyone who has ever watched The Polar Express or any other Rober Zemeckis directed motion capture animation will already be familiar. People generally feel more and more comfortable around robots that look more human, until, at a critical juncture when the inanimate object looks 'almost human' trust falls off a cliff. Meadows describes the feeling as being particularly palpable when he visited a Japanese engineering and programming genius working on a life-like female droid - the more he looked at the lack of imperfection in her skin, the dullness in her eyes, the lack of vitality in her expression, the stranger he felt.

Meadows also skewers the 'openess and transparency' arguments put forward by large technology companies such as Facebook, Apple and Google, which already collect massive amounts of data about their users via websites and mobile gadgets. Given that I share much of his discomfort at the idea of privacy controls being relaxed and so-called 'openess' becoming the norm, it is heartening to read an articulate privacy advocate who is not a Luddite and who does not succumb to paranoid distortions. 'And why don't we like strangers looking through our emails or chat logs?' Meadows asks rhetorically. 'Because the more information someone has about you, the more control they have over you. And to once again refer to Linda Stone's term , "understanding workers", it is worth noting that as information becomes knowledge, knowledge becomes understanding, your behaviour and your thoughts quickly become predictable. Privacy is liberty.' Spot on!

There is a footnote to a fascinating research paper that details how 'sentiment analysis' on Twitter can already predict likely box office outcomes with (some might think) alarming accuracy. An over active imagination with a tendency towards the dystopian could run away with itself - what else might such massive agglomerations of data help to predict? Who is making those kinds of inquiries? Might the same system be used to 'push' people towards a desired outcome without their knowledge or consent?

The possibilities are exhilarating - both amazing and frightening to think on't.

Meadows own work with artificial personalities informs his thoughts about the implications of mass data collection by corporations with clarity and insight. Like a lot of good storytellers, he posits questions but does not tell anyone what to think. Who owns the personality profiles that companies compile? What should companies and individuals be able to use those profiles for? - so-called 'sock puppets' (multiple accounts controlled by a single operator) deployed to influence debate and promote a particular agenda online - on forums and in comments sections - are already known about and being reported upon. With whom should such valuable data be shared? The questions are massive and, like Blade Runner, seem to approach fundamental questions about identity and what makes us who we are.

There is a lot of detailed research in the book but, as the allusions to fictional characters make clear – more especially the references to pop culture and comic book icons – We, Robot is not and was never intended to be a scientifically rigorous portrait of the current state-of-the art. Nor does it claim to be – in a scene that seems to directly reference William Gibson's seminal cyberpunk classic Neuromancer, Meadows describes stepping out of his hotel in downtown Tokyo, high on prescription medication, trying, in the hazy neon night time, to feel the signs of intermediation between human and robot reality. In my view, a book written by someone who knows a little bit more than the average bear about the kinds of systems that might be heading our way and which takes a responsible view of big issues to do with privacy, the psychology of uncanny valleys and the rights to ownership of our personalities, when used for marketing purposes, would be very welcome. As people try to understand the complicated relationships that are emerging between internet services and their users, these issues are only going to become more important.

Personally, I would love to read a more earnest account of Meadows work in artificial intelligence laboratories, performing semantic analysis and using that data to create natural language processing algorithms. The anecdote about his work heading a team of programmers and linguistic specialists in the design of some 'very intense software' which was able to scrape personalities and regurgitate them to a user - the team 'scraped' Arnold Schwarzenegger interviews and 'When the system was asked what it thought of gay marriage it replied, “Gay marriage should be between a man and a woman and if you ask me again I will make you do 500 push-ups.”' - is terrific. But I wanted more.

We, Robot is half technology dilemmas, half joyful romp through the maelstrom of modern robot tech. Still, if you want an intelligent guide to what technologists dream about, which touches upon, but does not get bogged down in, ethical issues, this one is highly recommended.

Now, you will have to excuse me, I am off to wrestle sharks off the coast of Western Australia, but not before I build a robot Mark Zuckerberg.lunarpark.blogspot.com - Do Androids Dream of Electric Sheep - Keyword description


Blogger Moonshynn said...

The book sounds fascinatingly relevant and therefore a compelling read. I'll be ordering my copy presently. Solid review.

10:50 PM

Blogger Eric Payne said...

Thanks. I hope you enjoy it as much as I did.

10:19 AM


Post a Comment

<< Home