Monday, 16 April 2018

A little knowledge is a dangerous thing...

...almost none is a disaster. Here's the House of Lords on Artificial Intelligence:
“These principles do come to life a little bit when you think about the Cambridge Analytica situation,” he told the Guardian. “Whether or not the data analytics they carried out was actually using AI … It gives an example of where it’s important that we do have strong intelligibility of what the hell is going on with our data.”
The HoL (or one of its committees) has published a call for regulation of AI because of scary foreign monopolies:
In a wide-ranging report, the committee has identified a number of threats that mismanagement of AI could bring to Britain. One concern is of the creation of “data monopolies”, large multinational companies – generally American or Chinese, with Facebook, Google and Tencent all named as examples – with such a grip on the collection of data that they can build better AI than anyone else, enhancing their grip on the data sources and creating a virtuous cycle that renders smaller companies and nations unable to compete.
It's not at all clear as to whether the committee's concern is data collection and use (for which we have the new GDPR regulations and the Information Commissioner) or artificial intelligence. They produce a vague set of 'principles' (including one lifted straight from 'I Robot' - one wonders whether they read the book) and then present the standard response of fussbuckets:
“Of course, if in due course people are not observing these ethical principles and the regulator thinks that their powers are inadequate, then there may be a time down the track that we need to rethink this.”
Ah. "Nice AI you have there, would like to see it damaged"!

Judging from the principles, the nonsense about Cambridge Analytica and the threat of unspecified regulation, what we have here is the classic approach of the ignorant - "I've never tried it but I don't like it" - combined with regulatory authority - "we don't know what it does so we'd better stop it just in case". And if you don't think they're ignorant, suck on this...
“We want there to be an open market in AI, basically, and if all that happens is we get five or six major AI systems and you have to belong to one of them in order to survive in the modern world, well, that would be something that we don’t want to see.”

No comments: