Facebook iconTwitter icon
B2B-Network:
Artificial Intelligence Failing To Deal With Transparency Challenges?
Artificial Intelligence Failing To Deal With Transparency Challenges?
Time icon 2 May 2017, 9:54 am

Many breakthroughs have been made in the field of artificial intelligence and these changes, despite coming with benefits, also availed new challenges. As artificial intelligence advances into higher levels, it has become more challenging to understand the inner workings of these algorithms. One of the reasons this happens is because most companies that design AI systems don’t offer allowance for the scrutiny of the algorithms. This coupled with the increasing complexity of AI presents a major transparency nightmare that experts are yet to resolve. 

 

AI algorithms are more efficient and safer than humans

By a large extent, AI algorithms perform tasks more efficiently than humans. Take the example of self driving cars, which are driven by machine learning algorithms; the success of these cars will reduce accidents by up to 90 percent. With features like predictive maintenance, it becomes easy to detect signs of wear and tear in a manner that humans cannot do and this prevents disasters on roads. 

 

Algorithms are not always comprehensible

But with all these benefits, there is still a thing about artificial intelligence that even engineers seem not to acknowledge: Sometimes it’s not possible to understand why an AI algorithm makes a certain decision. That AI algorithms are given full control to decide and do what they want is a worrisome issue since you are not certain what to expect from them. Manufacturers will face hard times explaining an event that comes as a result of many complexities in the decision making parameters of the algorithm. 

It is necessary to think about this because in near future, AI algorithms will be used in places like hospitals to offer treatment to patients. Sometimes the algorithm could  make decisions that are not in line with what it is expected to do, and here is where the problem lies: Because understanding the reason why the AI algorithm chose such action is a big challenge. There is also the issue of ethics and best practices that has to be hammered out before this field advances to greater levels. 

Loader