Code Monkey Island

Code Monkey Island

Friday, 17 June 2016

Google's A.I. Kill Switch

If you're one of the 36% of people who believe that artificial intelligence (AI) poses a threat to humanity, Google are coming to your rescue by developing a 'kill switch' to hold back the inevitable AI apocalypse.

Back in January I attended a conference on emerging technologies in legal services and there were several sessions dedicated to the use of AI. The benefits to those businesses starting to leverage AI were clear, with software learning to conduct the more mundane and repetitive tasks that would normally take an employee days or weeks to complete.

But virtually every speaker had to start with a myth-busting intro as it was clear that as soon as they mentioned artificial intelligence the audience would just be sat thinking about Terminator endoskeletons marching across a field of human skulls. In fact many of the speakers were on a rebranding exercise to get people on the term 'Cognitive Computing' and distancing the technology from the more common AI connotations.

Terminator 2: Judgment Day Robot
What 36% of people think the next Android phone will look like

Google have been working on a more practical solution to the public's AI concerns by implementing a process known as 'safe interruptibility', where learning systems will always be able to stopped by a human overseer. Crucially the process also ensures that the AI doesn't learn about this interruption capability and then takes steps to avoid it.
"Safe interruptibility can be useful to take control of a robot that is misbehaving… take it out of a delicate situation, or even to temporarily use it to achieve a task it did not learn to perform." - Stuart Armstrong
The Google engineers gave an example where an AI was taught to play Tetris and actually learnt to pause a game forever to avoid losing. This behaviour wasn't exactly a WarGames scenario but still an indication of how AI may solve problems using unconventional means.

No doubt an entire sector will develop around AI safety technology and as public concerns grow so will the industry. In the future we may have to look forward to having our AI 'anti-virus' keeping our AI personal assistant in check.

I think we're a long way off needing to seriously worry about AI bringing about the end of the world, but in the shorter term it may bring about the end of some jobs.

Wednesday, 1 June 2016

Three, 2, 1 ... Ads off!

15th June looks to be an interesting day for advertisers with mobile provider Three switching on their network wide ad-blocking for 24 hours.

Three's website advert blocking technology promises to 'get rid of the ads you don't want to see, while keeping the stuff you're really interested in' and who wouldn't want that? But as Three carry 45% of the UK's mobile Internet traffic some brands may not be too happy with their ads falling into their 'you don't want to see' category.

VW camper van covered in badges with speech bubble Can we squeeze a few more ads in?

No such thing as a free lunch

Seasoned Internet users realise that if any online service is 'free' then it's you (or more accurately your personal data) that's the product. Google, Facebook and Twitter aren't charging you a monthly subscription because their revenue is coming from advertisers paying them to show all those ads and promoted posts.

So individual users installing ad-blocking apps has always jarred with this model, however a mobile carrier implementing ad-blocking at network level is something else. In fact the ISBA (the trade body representing advertisers) has labelled Three's plans as nothing short of 'disastrous for the mobile advertising ecosystem'.

A new version of the Internet

Most users accept advertising as a part of their web experience and the Internet would be a very costly place without them. But anyone who's visited a commercial news site recently would be hard pushed to deny that the number of adverts has grown significantly. Sponsored stories, banner ads and everyone's favourite click bait articles often fill up the browser and bury the content you actually came to view.

Website with lots of adverts

Ever since my good old Nokia 7110 days I've treated data allowance as a limit not a target, conditioned over many years to hold off watching YouTube etc. until I'm on a Wi-Fi hotspot. So rage naturally ensues when scrolling down a website only to hit a rollover advert with a four minute HD trailer for something like Transformers vs. Ninja Turtles* that happily munches through my mobile data without a second thought.

If Three are able to successfully filter out unwanted ads then it may push the various advertising platforms to drop poor quality and intrusive advertising, which can only be good for consumers. But the more cynical say Three's end game is to ultimately start charging advertisers to push ads on their mobile network as an additional 'ad serving tax'.

I'm of the view that well-designed, targeted and infrequent ads are generally good thing. In the long run a blanket blocking of ads would only harm content creators and force consumers to pay by other more direct means.

* If Michael Bay is reading're welcome

Wednesday, 25 May 2016

LinkedIn Hacked - How safe is your uninspiring employment history?

It probably doesn't keep you up at night worrying that your LinkedIn profile has been compromised. After all, a hacker finding out that you're a 'results-orientated team player with excellent interpersonal skills' may not be the biggest invasion of your privacy.

But as detail of the latest hack materialises, are there wider concerns beyond someone maliciously accepting all your pending connection requests?

To recap, back in 2012 LinkedIn security was breached resulting in 6.5 million passwords being dumped onto a hacker forum. Shortly afterwards LinkedIn identified and disabled the affected accounts and added new security measures to their sign in process. Fast forward to May 2016 and another batch of account details from the same hack are now up for sale, with the hackers following up their 6.5 million password appetiser with a 117 million main course.

While the stolen passwords were encrypted it's taken just 72 hours for various groups to crack 90% of them and in the process reveal some frightening insights into people's password behaviour.

  1. 123456 - is the most common password used 1,135,936 times
  2. linkedin - comes in second with 207,488
  3. password - the security stalwart has surprisingly been edged into third place with 188,380
  4. 123456789 - is next with 149,916, obviously the password of choice for those who think using 123456 is bonkers

An honourable mention must also go to the password maggie which somehow gets into the top 20 after being used 30,972 times. Perhaps going to show just how many Conservative MPs and councillors are lurking in your business connections.

Your password is bad and you should feel bad - Zoidberg shouting meme

LinkedIn are correctly reminding users of its security best practices which contains helpful guidance on setting up two-step verification for signing in.

I must admit I've always been hesitant to attach my mobile number to LinkedIn for the fear of being bombarded with notifications. However two-step verification seems to work well and my mobile hasn't (yet) been flooded with inspirational leadership quotes....