Accessability Links

AI development raises new questions on ethics

From SIRI and Alexa to self-driving cars and drones, from deciding whether we can have a bank loan, or are invited to a job interview, artificial intelligence (AI) is everywhere. And it raises many questions – particularly when it comes to ethics. While AI-powered robots are typically programmed to do a specific task or reach a particular goal, the robots will make their own decisions to reach the goal. It’s these decisions – that are made on their own – that can cause issues, because the programmers can’t always take all of the robot’s possible choices into account. For example, if a driverless car is on the road when a dog unexpectedly crosses the road – should the car run over the dog or swerve and risk injury to the car’s occupants? Would it be the same answer for a cat? A squirrel? Or, another human being? What about five human beings?

Ethical concerns about this fast-moving industry have focused on AI in being used as deadly military drones, or that it could summon the end of global financial systems. And as autonomous driving looks to replace drivers, there’s anxiety around unemployment. There are lots of examples of AI reinforcing what it’s learnt from real-world data, including racial or gender bias. Amazon, for example, developed a recruiting tool for identifying software engineers to hire, but discriminated against women. In 2016, ProPublica developed a system to predict the likelihood of criminals reoffending – and was found biased against black people. The problem here is that decisions are not easily accessed by humans so it’s not possible to look at where it started to go wrong.

So, despite advances in the technology, nothing will solve the fundamental problem of AI – and that’s that a well-designed, perhaps even perfect, algorithm will make decision based on a flawed and unpredictable world. Compared to human intelligence, which is robust, capable to learning something in one context and applying it to another, and uses common sense, AI appears fragile. 

Read our related blog: How will AI impact software development

Add new comment

Meet the team

Back to Top