Having spent two days at the excellent
AI Congress, I've come away not really sure if I've grasped AI, or whether I'm even more confused. Depending on who was talking, AI is either snake oil, mimicry of animal psychology or advanced machine learning with a dash of deep learning and neural networks sprinkled on top, executed by data scientists fresh out of college on £100k a year salaries and problems working in teams.
First impressions first; like a lot of tech, the visualisation or branding of 'AI' leaves a hell of a lot to be desired. There was a start-up village of stall holders, complete with bloodless merchandising, and names like CloudText, Botweb and ClickKnob (ok I made those up), and the branding of the event itself featured what looked like C-3PO's withered robot hand holding an eye ball. There was even a little robot buzzing around on day one. What did it look like? You guessed it, the most anonymous looking white plastic-coated shell of a thing on wheels, that you might stand on to check your weight.
My conclusion from this, is that not enough imagination or structure is being applied to the benefits of AI. No wonder there's no clear definition of it. The only clear definitions to be seen were the elegant data science models and technical architecture beaming off screens in presentations (Gousto's one in particular was a thing of beauty, guaranteed to get data architects all worked up and rubbing their thighs, but leaving the rest of us fairly perplexed).
But pull away the semantic complexity (this itself, the subject of an excellent hypothesis: voice is the new UI), and you're left with some extremely sensible advice:-
Focus on problem definition, and use machine learning to make better business decisions
Don't worry for now about using AI as an agent for execution
Do not underestimate the importance of good data logging and cleaning
Keep human beings close to the machines at all times, as things will quickly go wrong when anomalies occur (real life has a habit of being chaotic)
Create a culture that encourages both data scientists and business managers to work alongside each other, with the space to deliver a well defined, quality outcome
One thing's for sure, AI is certainly 'hot' right now, and very much at the peak of inflated expectations, as analysis of M&A and even historical AI conference attendance figures demonstrated. Which means that if you're a Data Scientist who demonstrates even a passing interest in business, and you can look someone in the eye and train them up to be as good as you, you'll be pulling in the big bucks. Oh, at least until Data Science Automation lands, but that's another story …