It’s Time To Stop Mixing Up AI and Machine Learning

There is a lot of buzz/hype around artificial intelligence (AI) and machine learning (ML) – some of it for good reason. With DeepMind beating the world’s number one player at Go, and Netflix utilising machine learning to recommend shows to users, AI and ML has many industries excited (and worried), about its transformative implications for how we work.

It has various members of the C-Suite getting excited about the potential revenue streams and new business opportunities. Yet, confusion reigns – artificial intelligence and machine learning mean the same thing for a lot of people/too many people.

This is simply not the case. This lack of understanding is the first fundamental hurdle organisations are facing around artificial intelligence and machine learning. After all, how are you supposed to extract value if they are used interchangeably when they represent different things?

Organisations must know the difference

As Bernard Marr states: AI is the broader concept of machines being able to carry out tasks in a way we’d consider ‘smart’ (for example, recognising a face in an image or reasoning an answer based on a question). Machine learning is an AI application based around an idea that we should be able to give machines access to data and let them learn, predict and classify based on the input they are given. It’s similar, but it’s not the same – and it’s making some in the tech industry look foolish!

Buzzwords are a problem for the enterprise. They become so overused they become meaningless, leading to miscommunication and misunderstanding – and possibly to wasted investment. It leads to an unfounded belief that AI and ML can be used to solve any business case or problem. The over-hype that exists around AI and ML right now will mean that a great number of enterprises will increase their odds of becoming another failure statistic as they rush to figure out how to use shiny new toys without putting the right platforms and procedures in place based on real understanding.

It will be hard to have an analytical culture and really compete on data science without a growing awareness of what tools are what, in the marketplace.

It’s hard to get the shiny toys to work

The big ‘FANG’ companies (Facebook-Amazon-Netflix-Google) understand the complexities of getting it right. They possess understanding that by leveraging the data deluge carefully, by laying the right groundwork, they start with a strong foundation. The data they have is in the right place, well kept, understood and curated – meaning a solid foundation is in place for any kind of transformation or experimental project. That’s no small part of the reasons behind their growth and dominance in tech.

However, the C-Suite must note that even the most well-resourced, and most ‘data ready’ of enterprises are not immune to artificial intelligence mistakes. An unfortunate demonstration of this can be seen with Tay, Microsoft’s chatbot. Originally designed as an experiment to engage with real people using light conversation, Tay started well… doing just this. Unfortunately, this descended rapidly into chaos – with Tay pouring out inflammatory, offensive remarks from the Twitter account, based on nasty human input.

Microsoft’s error was a demonstration of woeful utopian naivety, based on assumptions that everyone is as nice to each other online as they are within the halls of Redmond. Simply not at all preparing its chatbot for the worst elements of human society allowed it to learn ad behaviour in a public setting.

Microsoft is not the only one to struggle. Facebook also recently had to scale back their chatbot projects, citing a 70% failure rate. The mistakes these companies made really demonstrates the various challenges AI poses. AI is hard. With the vast majority of the industry still not really really understanding the nuances of deep learning networks, disaster could ensue, with worse results than the bigger players, who already have the data groundwork in place.

Lay the groundwork, then focus it

It is critical that organisations don’t jump before they can run. A good example of this is another inescapable enterprise buzzword – The Internet of Things (IoT).

Research conducted by Cisco, around the failure rates of IoT projects, discovering that three out four IoT projects will fail. Cisco cited challenges such as the time taken to complete, limited internal expertise, quality of data, and integration across teams. Similar challenges to many kinds of new transformation that require a strong data element to succeed, including AI and ML.

The success of any transformation or experimental projects hinges on the ability to garner smart data insights to improve processes. However, without having the right data tools, or the correct data culture to help encourage this, you are on a ‘hiding to nothing’, as the saying goes.

The take-way is that AI is ‘hard’ and the industry still doesn’t really understand the nuances of Deep Learning networks to apply them everywhere they’re needed. Secondly, AI (as ML before it, and descriptive analytics before that) needs to be focused on a tight business problem. It’s not a one-size fits all solution and requires situational awareness and a laser-sharp attention to the customer/end-user need in order to succeed.

But that just means there is an opportunity out there to be seized by those prepared to get to the heart of the problem and use the world of analytics smartly. Creating a data culture, and having knowledge workers take the lead in transforming the world using their best resource to make intelligent and positive change – data. It’s a wild ride!