AI will create 2.3 million jobs in 2020, while eliminating 1.8 million
Artificial intelligence is more than a buzzword today. As the technology matures, and industry and governments increasingly look at adopting AI, there is also a need to have a conversation about the ethics of AI systems.
Pune-based Persistent Systems is tackling some of these issues as they work with clients and within the company to understand how and what kind of ethics-related questions are likely to impact the development of AI.
The question of ethics in AI is broadly centred at two levels — robo ethics, which involves the ethical or moral side of people designing AI systems and machine ethics, which deals with the question of whether a thinking machine is ethical, explained Rashmi Tambe, a technology evangelist at Persistent.
As an example, she cites an algorithm in the US, developed to predict crime rates was found to be biased against the black population. The reason, it was determined, was that the people or systems generating or feeding the data to train the algorithm on, were consciously or unconsciously biased against blacks and that crept into the algorithm as well.
AI has graduated from simple automation to systems like neural networks that think like humans and are increasingly being tested for deployment in various fields such as healthcare, retail and so on.
“Some other ethical issues are impact on human behaviour — technology addiction, reliance on Google or some other technology systems. It is an ethical issue — at what point you stop, and at what point you unplug. It is going to become bothersome at some point, but I believe there are ways to handle it,” she said.
With AI, there is also the constant perception battle that AI will eliminate jobs. While doomsday predictions abound, several reports have pointed to a changing nature of jobs scenario in the near future.
According to research firm Gartner, AI will create 2.3 million jobs in 2020, while eliminating 1.8 million. "Unfortunately, most calamitous warnings of job losses confuse AI with automation — that overshadows the greatest AI benefit — AI augmentation — a combination of human and artificial intelligence, where both complement each other," Gartner analyst Svetlana Sicular had said in December.
Persistent’s Tambe also said customers using AI are also asking questions about the ethical issues around AI.
“What happens if bots go haywire? If there is a mistake, in a robo advisory, who gets sued? Who gets the liability? These kind of questions are coming up. We are firming some of these guidelines and working with industry and auditing firms and those who do consulting and auditing around it to help our customers,” she said.
Another question that arises then is whether it is even possible to have a completely unbiased AI system?
“Having very diverse data sets is the only way to handle bias,” said Tambe. “Machines start understanding patterns by looking at the data… So, data scientist has to check for balance of data,” she added.
For example, in a breast cancer detection system, it is not just enough to have the data of women of various ages, race, geographical location and so on. If there is not data about say, pregnant women, the system will have no conclusive inference on how breast cancer impacts this set of people.
There is however, a need to start addressing some of these issues right away. And until there are laws or industry standards to deal with these issues, it will remain a grey area.