Answer by Ian Cosgrove:
Convolutional Neural Networks. Instead of training a neural network to recognise what an object is, CNN uses multiple neural networks. Each one trained to recognise individual features that add up to one overall identification.
Pointy ears+whiskers+sits on keyboard while you're working = Cat
CNNs are what enable Google Now and recent improvements in speech and character recognition
Cognitive Systems, like IBMs Watson. Provides some of the features you would traditionally look at expert systems for but rules don't need to be hard coded. Cognitive systems excel at big data and pattern recognition to provide analysis decision support systems, Watson assists in diagnosis and selection of cancer treatment.
Blended Datasets, combination of multiple indirectly related datasets. Looking at an upcoming predicted heatwave and realtime power usage you can predict how all of those air conditioners kicking on will affect the grid. Blend in travel data like booked airplane tickets coming in (coastal region) or going out (cities) and you have a better idea again.
Platforms, Hadoop and NoSql DBs are changing things and there's still a far way to go. Install Ubuntu server 13 and MongoDB now.
Security, how do you keep all that sensitive real world, real people safe and secure. Then how do you monetize it?
Internet of Things:
Sensors, IoT is not just an internet enabled refridgerator. Realworld sensors will allow autonomic systems to be aware of what is happening in the real world in real time, understand it and take action in real time. From Nest thermostats to smart meters to the whole SmartGrid the IoT can enable massive power/money/carbon savings.
Jobs for makers, the IT industry has been moving away from embedded and low level/hardware developers for years leaving that area up to hobbyists, hackerspaces and the maker community. That expertise is needed more than ever so find your local hackerspace and see when they are holding their next open day/workshops and see what this sensor/arduino/raspberry pi business is all about.
Human Sensors, anyone with a smart phone can be considered a node in the IoT, people contribute real world data as it happens about weather and traffic events, traffic routes etc and social contribution apps will only get bigger thanks to successes like Waze
Cloud and Web Services:
What? Aren't they the buzzwords of 2-14 years ago? Yes but what do you think enables all of these IoT, BD, AI things to communicate. Services will continue to become more modular and composable until AI can identify your needs and weave a custom solution out of the cloud. When things go wrong you're going to need an increasingly self diagnosing/healing interweb and that's going to need some new types of magic.
Object Oriented languages were a great step up, the problem is that they're too good. They don't speak concurrently and they lack the correct verbage for building big big systems of the future but they allow us to build things that work so we put up with them to solve problems they were never intended for. I'm thinking something based on Set theory..
Wearables and blended reality:
Yea confidence in the market took a knock after the Nike Fuelband situation but how many times did tablets fail before the iPad launched?
I'm not a huge fan of augmented reality but come on, Word Lens did it right. Just a pity Glass didn't go far enough. Then on the other end of the spectrum you have Oculus Rift.
Add in some big data from the IoT categorized by CNNs and analyzed by cognitive system and you have just elevated the capabilities of the human mind.
So aspects of all of these things exist on the fringes, you may have heard of some or all of them already but they are still cutting into new frontiers and exposing new challenges. It's where these trends will be combining that real fringe opportunities are emerging.