AI won’t absolve you

2016 will be the year of AI or so lots of articles, consultants, conference talks and tweets tell you almost robotically. I don’t disagree that the fields related to AI are starting to mature but we are a long way off from true general AI. Never the less this year will see a lot of people in a lot of organisations thinking they need to get some AI projects on the go. I worry for a lot of them. AI will not absolve you from design, effort, communication etc. etc.

There are a lot of fields underneath the banner of AI.

AI ecosystem

My worry is that artificial intelligence and machine learning as brands conjure up fanciful images of answers being produced as if by magic by these mega software and hardware beasts. Lots of things being conflated underneath the brand of AI. It helps fuel the hype it can also deform the reality.

AI will not be the answer to all of your problems. You can’t just AI away design problems. It might fix some issues and help you do some things you never could but it certainly won’t fix them all and it will certainly throw up some interesting new complex ones.

And talking of complex, complex software requires ever more complex skilled humans to understand and implement well. You cant just take all that lovely big data you have been hoarding for years and doing nothing of note with and fire it into these things and expect magic to pop out?

Without getting into the whole complex issue of bias built into the systems by the humans that design them, what does fake/machine learned empathy look like? Then again what is human empathy really?

There is definitely phenomenal potential in AI advances but it is still in its infancy and infants need a lot of adult supervision. They are capable of brilliance in between tantrum laden meltdown, soiled underwear, and refusal to eat what you try and feed them sometimes leading to spectacular vomiting.

“The business plans of the next 10,000 startups are easy to forecast: Take X and add AI. This is a big deal, and now it’s here.”?—?Kevin Kelly

Before buying into the hype get some of your best and brightest data science nerds and architects and designers to help you get a real insight into what will be involved, what outcomes you can realistically expect and ensure you treat it as an experiment rather than a sure thing. Like anything the more you put into it the more you will get out. AI will not absolve you of defining the problem you are trying to solve.

The people and companies that succeed in using AI well will no doubt be similar ones that designed mobile interfaces well, have simple and clear services elsewhere, get real data from their own data today. It just comes down to the old classic of spend more time than is reasonable defining the problem/making something simple. Spend less time than is reasonable hoping the robots will fix it for you is a recipe for disaster.

Leave a Reply