“AI researchers are warning developers to focus more on how and why a system produces certain results than the fact that the system can accurately and rapidly produce them.”
That is an ominous start to this article from Vice. The steps we’ve seen from AI in recent years are pretty incredible – neural networks, natural language processing – these terms are becoming more in our vocabulary as AI moves from the theoretical to the real world use case. But just as the results from AI get the tech world excited, how they get those results leaves researchers baffled.
Autonomous cars, content recommendations, customer service chatbots, identifying new drugs and diagnosing diseases. These are just some of the real world applications that are emerging. Chances are, without even knowing it, you’ve likely interacted with AI or benefited from the work of AI in some way, shape or form (as much as you can say that getting recommended the best TikTok or Instagram Reel is ‘benefiting’).
But these ‘black box’ systems – where we get an output but don’t understand how the system got there, will become a problem. The challenge here is two-fold. Firstly, if we don’t know how a system is getting to a certain output or answer, we will be less certain if that answer is correct. Basically, if you’re asking an AI system to diagnose a disease you don’t just want the answer, you want the steps it took so you can check its work.
The second challenge relates to adjusting or fixing these systems. As they get more complex and trained on more data, it will be harder to understand if there’s any issues in how they’re being trained if we don’t have a better understanding of how they work. An example that has popped up recently has been AI mimicking racial or gender biases. Having a better understanding of how these systems go about their tasks, will allow researchers to better understand when there are adjustments that need to be made.
The age of artificial intelligence is here. The use cases may only be novel (like AI art) or seemingly insignificant (like social media recommendations) but that is going to change. And according to the experts, change quickly. So these are the conversations that need to be had today.
This is an excerpt from our Thought Starters email. Once a week we send you 5 interesting articles that have caught our attention, to get you thinking. No spam, we guarantee.