Having joined the product management team at Futureheads back in August with an appreciation for emerging technologies, I quickly became interested in the field of machine learning and its practical applications. I thought I'd write my first post on machine learning as well as UCD 2016 (the two very much overlapped this year).
Machine learning is generally a way of making a system better with examples, usually in a supervised or unsupervised manner. It is being used in almost every field – fighting spam emails, credit card fraud, face detection and recognition, speech recognition. Google is also a machine learning company. And if you haven’t heard of IBM Watson yet then you are in for a treat - software built on the Watson platform is now being used in the legal field! It’s definitely a trend that’s teetering at the top of the Gartner curve.
The machine learning/AI hype was palpable at this year’s UCD 2016 that I attended recently. This year UCD’s theme was ‘Being Human’ and it highlighted the philosophical tension between humanity and science in the digital age. Hardly surprising that machine learning and AI were recurring topics throughout the day.
But how are we going to use this technology and the developments in AI and Machine Learning? Will we fund and build products that solve actual human problems or unintentionally be the seed fund and technology incubator for Skynet and bring about the rise of the machines!?
Probably not, but emerging skills shortages are becoming apparent. According to a McKinsey study, it's estimated that by 2018 the demand for data scientists with a computer science or machine learning background will outweigh supply by 50%. Previously, companies aimed to be technology driven or product-driven in the digital age, now companies like Google, Apple, Facebook and Amazon (GAFA) are aiming to build machine learning into all their products and be machine learning first.
The problem with machine learning is not building machine learning algorithms, which I’m told by various thought leaders in the area including Pete Trainor (Nexus) and Devika Thapar (IBM) is the easy part. AI and machine learning algorithms learn (as their name suggests) and are not taught. This is not a problem for companies that have millions (or billions) of users. They have a vast pool of data to train their machine learning algorithms. Great for them. The largest companies have the richest data to build the most accurate algorithms.
But what about start-ups? Where can they access their data? How can they get users to engage with their products to generate new data? Google already has the cloud machine learning, with pre-trained machine learning models that can be tailored to suit needs, which start-ups can access. However, this still means that the largest companies own the monopoly on data.
It will be interesting to see the effect on the product management market. The role of the product manager requires the spinning of many plates. This technology shift suggests a need to be more and more comfortable with working with huge amounts of data, supported by data scientists to build more user-centric and personalised experiences. The Economist states that 40% of the top data scientists in the field are now associated with companies rather than research institutes/universities.
The fact that UCD is a user-focused conference inevitably led to the debate over the way in which to use technology and machine learning to enhance human experience, not replace us. This transition into the Third Computer Age is no different to any other revolution. Certain skills are outdated but open the door for humans to work on interesting problems and not mundane, repetitive and labour intensive tasks.
An example of practical use would be to cut out the monotony of everyday tasks and to enhance human skills rather than to replace. Take booking a holiday for instance. As Dave Coplin from Microsoft puts it – ‘going on a holiday involves jumping through various technological, time-consuming hoops’. First, you have to spend an evening trawling through search engines to find a destination. Great you’ve found one. Now you spend another evening trawling through price comparison sites, sky scanner and third party travel sites to book your flights. Great, your fights are booked. And finally a third-night trawling through search engines and price comparison sites to find a hotel.
With all this data, machine learning could learn what you, person X, wants from a holiday – cut out all the laborious middle bit, negotiate and book your holiday. It’s these kinds of technologies that add value to human existence rather than detract from it.
IDC predict that by 2018 around half of the developers will entwine cognitive systems in their apps, which will be focused around machine learning and AI technologies.
The path is clear; products and services that are based on data and machine learning mean competition for certain skill sets will grow globally. The advantages are to those companies with the deepest pockets, to acquire the best/most relevant talent or the ability to acquire companies that already employ those skillsets.