To AI or Not to AI

There has been a lot of hype about AI* lately. Every other day there is news about a major tech giant announcing it as the next big thing, another academic endorsing it to be the next ‘platform’. Another overnight success…this time over 60 years in the making! Many experts say this time it is for real.

Having experienced various waves in the last 20 years, there are some key factors which I think lend credence to ‘this time it is different’ scenario.

Data: One of the most fascinating stats I came across is that every two days we now generate more data than from the start of civilization till 2003. (source: Eric Schmidt, Tech Crunch). And that was in 2010 – today perhaps we do so in less than a day. This explosion in data has opened up the basic ability to provide enough input to the machine to learn, draw conclusions and become intelligent.

Processing power: The data explosion has been matched with a similar advancement in processing power. Today one microprocessor has 400,000 times greater computing power than the one in 1971, and an average smartphone is more powerful than a room sized supercomputer from the 80s (source: Economist). By some estimate there has been a trillion fold increase in computer performance from 1956 to 2016 (source: Expert Exchange). The machines are now capable of processing the raw material of ‘data’ thrown at them.

Underlying technology: In recent years, there has been a significant progress in ‘learning’ vs. ‘programing’ based algorithms. Shift from linear neural networks to multi-layered ones has provided a breakthrough. The biggest improvements have been made in machine learning, which enables various other applications/technologies…with driverless cars being a key example. Also, leading companies in the field like Google are making these technologies available on an open-source basis, helping trigger a wide ecosystem where developers use it to create applications and further enhance the technology itself. 

So we now have plentiful raw material and processing capacity, combined with tech advancements for compelling applications. All these developments are  feeding into user adoption. Interacting with Siri, Alexa and Cortana is becoming more commonplace and should soon become mainstream. There are a number of other devices/apps that are used by consumers today which rely on AI. A combination of tech development and user adoption could becoming the tipping point for AI to be the ‘platform’ on which interactions and services are delivered in future. 

All this has made AI a topic of discussion in several boards and innovation teams. Big tech giants (Google, Facebook, Apple, Microsoft, IBM etc.) are already committed to it and are investing billions. But like the internet itself, most businesses will need to have their AI strategy in place. The question is…do we start investing today, assuming it is the future? Or is it ahead of the curve? The timing question is always the hardest to answer. 

Lately, with my focus on the digital health I have been mulling the timing question. Healthcare is considered to be one of the key sectors with wide-ranging impact of AI. It has a myriad of possibilities.   

I believe AI will play an important role in healthcare. But without betting the farm on it, we are making measured progress by adopting some broad principles: 

Building AI readiness: we are putting a data strategy in place for our businesses, which makes us AI ready. This investment has several other benefits and, frankly, should be ‘hygiene’ for all modern businesses.

Leveraging existing products/offerings: we are scanning the ecosystem for ready applications focused on healthcare that we can use, rather than the proverbial ‘reinventing the wheel'. The ecosystem is getting richer with various start-ups as well as big players like IBM (Watson) having healthcare specific offering. A smart selection of offerings pieced together to deliver the business goals could be a time and cost efficient strategy.

Creating tangible impact: while AI has a very futuristic connotation, several applications are ready today and are rather modest and mundane. These include, for example, efficiency and cost reduction in customer services, increasing ‘adherence’ and preventing insurance fraud. Focusing on these should drive tangible returns and make the concept more 'real'.

Driving mini-transformations: and in the end, using all the above principles/initiatives to do small projects and create proof of concept. Irrespective of industry and technology, I believe driving mini-transformations is an effective way to do and learn, and get organisational buy-in.

Which brings me to the all important point of driving organisational change and building readiness. Needless to say, this is the single most important factor determining the success (or failure) of any initiative. We are not in the technology business and, as such, enthusiasm about new technologies is not natural. And there is always the sensitive man vs. machine argument. The hyperbole around machines replacing doctors doesn't help (I don't believe this is going to be the case anytime soon...but that is for a later discussion).  

We believe the gradual approach of building readiness, and driving tangible benefits through mini-transformations serves the dual purpose of driving organisational change and making a more efficient investment.

The future looks exciting and we are committed to play an active part in it. Perhaps the question is not ‘To AI or Not to AI’ but ‘How and When to AI’!

* Note: AI is a broad term, covering multiple branches and so could mean different things to different people. These include computer vision (image recognition), robotics, virtual reality and natural language processing. There are over 25 branches and several sub-branches listed in Wikipedia.