Machine horizon: Three of the most exciting AI developments

As a topic, AI’s popularity tends to rise and fall. Some have even argued that it rises at times when there’s a lull in tangible technological innovations, perhaps to give the geeks something to talk about.

In the past, AI has often been a blank sheet onto which we project our fears and fantasies of the future. We get excited about artificial intelligence or we get fearful about it, yet those hopes or threats have always felt abstract and faraway.

This time, though, it feels different. This time, we’re excited about AI because AI seems genuinely exciting. Deep neural networks are now allowing for proper machine learning (some might even say intelligence), rather than pre-programming or algorithmic variations. Google DeepMind’s victory over Go world champion Lee Sedol was quite rightly seen as a milestone, a validation of the neural network approach. The ancient Chinese game has more possible legal positions than the (estimated) number of molecules in the universe, which means a computer can’t be algorithmically programmed to make the right choice for any given state. What’s more, as Elon Musk pointed out, this milestone was a decade earlier than anyone expected — perhaps we’re much further along the artificial intelligence road than any of us thought.

Yet creating a timeline for AI’s progress is tricky — it’s important to remember that sudden leaps forward don’t necessarily indicate consistent growth. And while we wait to see what’s next, we have time to wrestle with the social, ethical and economic implications of machine learning. Yet I choose to remain ‘pragmatically positive’, not least because, as a software professional, I’ve always believed in the power of technology to change people’s lives for the better. And I’m genuinely excited about what’s happening.

1. Driverless cars

For instance, ‘driverless’ cars are about to appear on a road near you. Companies such as Tesla, Google, Mercedes and BMW are now leading the way in testing and releasing cars with self-driving features. It’s estimated that 10 million of them will be on the road by 2020 (a driver’s presence will still be needed though — we’re a long way away, legally at least, from the possibility of empty cars speeding down the highway). But safer, more relaxing, more productive car journeys are well on their way.

2. Data analysis

Then there’s the potential for deep neural AI networks to analyse massive data. Google DeepMind has already been (controversially) crunching the encrypted information of millions of UK NHS patients. The goal? To compare patient data and make health predictions based on it. If it works, then medicine could be transformed by a new capability for early diagnosis. And that’s just health — the implications are massive across all fields, from advertising to sports to finance.

3. Image recognition

A third development that particularly excites me (as it has potential to revolutionise my own industry of Digital Asset Management) is image recognition. Machine learning will transform how we collect, classify and categorise our current sandstorm of digital assets. Image recognition APIs that use AI to tag images have been around for a while but when Google Cloud Vision was launched at the end of last year, DAM vendors started to get excited. At present the results are mixed. However, I don’t think it will be too long before DAM systems have AI integrations capable of (accurately) categorising images by subject, type, person (facial recognition), activities, colours, objects, emotions and beyond.

But what does this mean for the user? Basically, that boring tasks like tagging and categorising will take place automatically, making adopting or switching DAM solutions considerably less painful. In the not-too-distant future, organisations will be able to take a ‘leave it to the machines’ approach when it comes to the daunting task of categorising huge and complex image libraries.

But this will only become a reality when the tagging APIs match human accuracy, and we’re not there yet. And there’s another complication. Organisations tend to have their own specific cultures, conventions and terminology when it comes to image classification. For image recognition to really be of value to DAM solutions, client-specific and culture-specific customisation is crucial. This is where feedback loops for the machine learning APIs will be essential.

Yet these are hurdles, not a brick walls, and I’m really excited about how machine learning is set to transform my industry. The upcoming impact of AI on DAM is an example of how it could set us free, at work and at home, from boring, mundane, repetitive tasks.

For better or for worse (and that’s still very much up for debate), ‘leave it to the machines’ could become our mantra for the future.

Martin Wilson is a commentator on technology trends and founder of digital asset management solution, Asset Bank.

Image Credit: Sergey Niven / Shutterstock