Machine learning is the latest field of computer science to make an impact on society. For those who don't know, the basic idea behind it is that machines can learn processes rather than having them programmed, allowing for significant automation in many roles. 2015 saw it begin the downward trajectory from the peak of inflated expectations on the Gartner Hype Cycle, heading into the trough of disillusionment, just in front of autonomous vehicles and behind wearable technology. It is yet to reach its potential, but the coming years are going to hopefully see this achieved.
To help this, the last 12 months have seen significant developments in how machine learning is being used and the use cases for it are likely to grow even further.
One of the reasons that machine learning could be seen as slowly developing, was that it was time consuming to create and effectively use. Thanks to the wider use of Apache Spark, this difficulty has been largely alleviated. With the speed in which information can be processed and used through the platform, it has become considerably easier to use for a variety of tasks.
One of the best known benefactors of this wider use of Spark has been IBM, who have utilized it in their Watson cognitive computing system. This is arguably the most praised example of machine learning, especially in natural language processing. The potential that the machine has in almost all industries will lead others to try and recreate it.
One of the keys to this emulation and development of the industry is going to be removing the barriers to entry that currently exist. This has already begun to some extent, with Amazon, Apple, Facebook, Google and Microsoft all investing heavily in the area. Facebook especially has been very active, opening up their powerful M personal assistant to some Bay Area users in August 2015 and even open sourcing their deep learning modules created for the Torch framework. It means that it will not only be a select few people within the huge tech companies developing on the platform, there could also be people experimenting with the technologies all over the world.
This has a clear benefit to the wider industry, allowing developers to work on new uses for the technology or even creating completely new companies that can effectively utilize machine learning technology.
There is little doubt that the technology has become more powerful in the last year, moving away from basic functional uses into far more complex realms. As more companies have joined the data revolution it has meant more data being collected and therefore more effective analysis. 2015 saw the start of third party data sharing for machine learning, with IBM striking a deal with the Weather Channel to help provide more data for their system.
Another element that has certainly helped this, not only in the last 12 months but across the last few years, has been the state and price of data storage. As the price of storage has decreased, it has allowed more to be collected, meaning that companies are less selective about what they collect. Combined with this is the growing ability to effectively store data in a useable way. It means that accessing this data becomes considerably simpler, allowing algorithms to utilize the data far more quickly and effectively.
However, the single biggest reason that machine learning is getting closer to its potential is because it needs to.
One of the most important uses for machine learning moving forward is going to be in cyber security, something that is high in almost every company's priorities. William Hill, one of the UK's largest betting companies, have turned to it already in their quest to stop hacks. Finbarr Joy, their CTO, claimed that cyber attacks are 'constant' and as such 'we're using a lot of technology to stop those patterns of behaviour. How do you spot a pattern that's different from the normal?' in an interview with Computing. They have adopted machine learning for this exact reason and Joy went on to say, 'When things are happening outside of the normal patterns, it makes those explicit then you can zero in on them. It's a good example of how a machine can learn what's normal and therefore can spot anomalies much more easily than a thousand people.’
With these hacks on the rise, the use of machine learning for cyber security will become a business critical idea, especially if more companies achieve success from using it. This kind of success with prompt further investment, increasing the speed of development and potentially seeing the technology reach its potential quicker than many would have predicted.