What if the Machine is not learning but manipulating
Machine Learning: A field of artificial intelligence that uses statistical techniques to give computer systems the ability to “learn” from data, without being explicitly programmed.
Thoughts on the direction the AI ML universe needs to take
What was once science fiction or wishful thinking, today is truism! 3 year olds have in their tiny palms handhelds as powerful as the supercomputers of yesteryears- the ones that propelled us into space or the ones that brought down Hiroshima and Nagasaki! So the title shouldn’t surprise as much as wake us up. This one isn’t a plot from a movie, rather ideas that may shape what lies ahead and for good. We manipulate in relationships, we manipulate our dog and yes our dog manipulates us for love. Rather our dog manipulates us way better than our spouses ever will.
In today’s tech age, we do not write programs anymore but Algorithms that not just program but have the ability to process, compute and learn from an incident and accommodate for every such incident and the constraint that came along the way. That is what we call Machine learning!
Like the brain which is trained to complete a task with the minimum effort required, the algorithms too are but an extension of our neurons and they do as asked (or as programmed) of them. The key focus of any algorithm therefore is to optimize and not ponder upon what ifs. I am okay with that, and that in itself renders machines the efficiency and speed we are so dazzled by. They follow a checklist to learn, they can not and will not seek the -so whats and now whats.
I am not a ‘naysayer’ or a pessimist and I am not taking this article onto that beaten path. I am rather largely asking some pertinent questions every AI, ML and deep learning enthusiast, programmer or visionary needs to ponder as we continue to make some giant strides in the space. These may well be some of the existential questions that may shape up the next generation of algorithms and bring a more heuristic perspective to the field.
1. What if the machine and its powering algorithms aren’t learning but manipulating us?
2. What if the algorithms aren’t taking inputs from our behaviour but piggybacking and feeding on our fears ?
3. How do we accommodate a higher sense of integrity into algorithms rather than just letting it be all outcome (revenue) based?
I leave everyone with two cases that we all may relate to and pit them against what a typical #ML algorithm does in these scenarios. I would also offer some hacks which as citizens of this era, we may use to nullify the mindnumbing algorithmization of our lives. But as programmers and developers of this industry, there isn’t escaping this through hacks, we will have to take the challenges head on and better empower our #AI universe.
Case 1: The confirmation bias at play
I am a Liberal American /or an immigrant in the US / or pro LGBT rights. I am also just beginning to understand American politics. I go to Facebook and like a couple of pages around these topics. I am thrown in some videos around these issues by the Facebook autoplay algorithm and I end up watching these short informative videos. By the end of my browsing session- while I have no idea of what Republican stands are, I have already ended up hating all republicans alike, not just Trump alone. Alternate scenario is that I am ‘Conservative’ fiscally, liberal in a few social aspects, conservative in some other with a belief in God. I end up again on the same train of (thoughts) videos and end up hating the LGBTQ community (for they are aparently against the will of Jesus), Obama and Obamacare (for its an anti capitalistic welfare scheme).
6 months down the line, I have a religion of my own. I am not a Christian anymore or Catholic or a Hindu or Muslim. Rather I am a liberal or conservative or an immigrant and every alternate view is a sin. There are many like me who I haven’t met, but they stand by me and my stands and some crazy bigots who abuse me as much on every social forum. Good I don’t have to meet them in real. I am attached more to stands I take or decided to take than real causes! That is billions of marketing dollars working its way to behavioural analytics nudging me to be who I really am. Every marginal inkling I had is now part of my belief system.
Case 2: The confirmation conundrum at play
I am an average built human with average features. My Instagram, Flipboard and Browser feeds keep flowing in beautiful celebrity pics that either make me feel I am thin or fat or worse, too thin or too fat! I just have an inkling at this stage that ‘somethings amiss that I must redress’. I ask around and I am told all’s well and I get on with life. But say I didn’t ask around for I am a person of science, and instead tried Googling my way to a definitive answer to this million-dollar question.
“Am I Fat?” are the three magical words I search for. Google very generously shares some of its biggest hits on Body Mass Index / BMI. Parity restores momentarily but there are some catchy hyperlinks around how to stay fit and lead a healthy life and shed that extra pound. Couple of articles and few days later, my YouTube has videos around weight loss, there is an Amazon ad that plays in between letting me know that there is a sale on all kinds of weight loss and gain products. One more click and I am led to believe there is something definitely wrong with me and I tussle with this idea for more days than I would have wanted to! Gladly Oprah’s ‘Weight Watchers’ and its consumer insights algorithms have made it possible for some very handy products that are custom tailored for me. What it does to my self-confidence is anyone’s guess.
There can be numerous such examples that affect our ability to think straight or our ability to think itself. That leads me to ask if ML algorithms as of today learn from us or mere feed on our vulnerabilities, insecurities, beliefs and biases and end up fortifying them?
There are simple hacks of not searching for our inklings outside the incognito mode, using browsers, devices and search engines from different universes (Apple, Google, Microsoft, Amazon) or following diverse pages with alternate perspectives on social media to restore some sanity and diverse line of thoughts. (Remember, in a Fox universe, America is great again, in CNBC’s and CNN’s America always was, but now isn’t. Wish things were as binary!)
The ones who are currently designing the algorithms that power the Deep learning and ML universe, hacks are no good! We have to think through the algorithms we set out to learn from us. For me every ML algorithm is a ‘baby’ needing to be raised and it needs care, caution and a higher sense of purpose. It’s as living as we are.
When Paulo Coelho in his book ‘The Alchemist’ writes “When you want something, all the universe conspires in helping you to achieve it” he wasn’t necessarily pushing ML but it has become all the more true today. If you search for something (could be anything- product, idea or a product idea), that something relentlessly starts searching for you.
#ArtificialIntelligence #AI #MachineLearning #ML #AIML #DeepLearning #Customization #Hyper #Personalization #Behavioral #Analytics