Artificial Intelligence and Fraud
Repurposing what I’d written sometime towards the end of 2018
So, yesterday in one of my chat groups, my friends and I were having a conversation on AI (Artificial Intelligence) and ML (Machine Learning) since another friend in that group was asked to learn ML for some project at his company. The very next day, to my utter shock (noooot), I bumped into quite a few articles on Twitter, FB, LinkedIn related to AI/ML.
A few of them were about products using AI, which obviously made me go WOW.
But another article, rather a post on LinkedIn got me thinking…
In that post, the writer says and I quote “This is what gives AI much of its power: It can discover connections in the data that would be more complicated or nuanced than a human would find. But this complexity also means that the reason the software reaches any particular conclusion is often largely opaque, even to its own creators…… It’s hard for humans to trust a system they can’t understand — and without trust…..”
This got me thinking especially “It’s hard for humans to trust a system they can’t understand…”.
Now let us take a scenario of an AI based system to tackle ad fraud. By definition, ML basically improves its decision making based on experiences and data. The assumption is that such a system will get smarter and will make sure that ad fraud is identified in a more efficient and reliable manner. But what if, and which is a big IF, such a system goes rogue? Since the systems learns about ad fraud through the data it is receiving, what if hidden connections are made within this ad fraud data to train the system so that it unlearns what ad fraud is and learns smarter way to let go of the ad fraud?
An extremely simple and non AI way to put this is how phone manufacturers cheat benchmark tests by running the phone in full performance mode as soon as it identifies an app as a bench-marking app.
Another recent example is of searching for the word “idiot” in Google and the results showing up images of Trump. While this may seem funny at a superficial level, isn’t it derogatory to the person and the position he holds on a global scale? (back then, he was the POTUS)
While these may not be a relevant example, I hope you guys are getting what I am trying to say here.
Like the quote above, it may discover data connections between these ad fraud data that would otherwise be complicated to a human and since its conclusion is largely opaque even to its creators, this can become a very dangerous proposition for the entire ecosystem.
The pace at which AI/ML is being put into everything is becoming interesting and scary. While everyone wants a piece of this, my question is who will put in the checks and balances and who will be responsible if things go south?
#technonolgy #machinelearning #fraud #artificialintelligence