The recent InterCon in Las Vegas featured a panel discussion titled "Is AI Really Beneficial For End Users?" Some key takeaways were that AI brings benefits by increasing productivity and assisting with problem solving, and that there is a need for governance and ethics in AI and a concern over bias in training datasets.
The panel was moderated by Rahul Bhardwaj, vice president of global privacy and data security at Duff & Phelp. Panelists included Areeya Lila, CEO and co-founder of VIEWN; Olivier Sosak, CEO at IT Y'ALL; and Ritu Malhotra, founder & CEO of dialoggBox. The panelists discussed questions posed by Bhardwaj and fielded a few more from the audience at the end of the session.
The first question from Bhardwaj was based on the session's title: what are the benefits of AI to an end user perspective? Sosak noted that AI has the potential to provide highly personalized experiences, for example, in tailoring educational programs for individual students, or for creating "virtual friends." Malhotra pointed out that AI is already being used in many products "under the covers," for example to enhance the quality of video streams. Moderator Bhardwaj also mentioned that in his experience in cybersecurity, AI is being used to defend against hackers; he also brought up applications of AI in farming, which improve crop yields and could improve the quality of life for many people by fighting hunger. Lila cautioned that regardless of the application, it is important to guard against training bias.
The conversation then turned to the panelists' top concerns around how AI is being used today. Lila repeated her concerns with bias in training data. However, she did express her belief that, contrary to a popular perception, AI would not eliminate jobs. Sosak agreed, pointing out that AI is often used for jobs that humans cannot do well, such as cybersecurity intrusion detection. He also noted that, like any tool, bad actors can use AI for bad purposes.
Bhardwaj then asked the panelists if they thought that concerns about AI could be addressed by governance or a code of ethics. All the panelists agreed. Bhardwaj compared AI to genetic engineering, noting a contrast between the restrictions on genetic engineering and the lack of constraints on AI research and development. Lila agreed, and brought up concerns about AI in weapons systems. Malhotra said that while many people are concerned with the use of biased training data, the objective function or goal that the AI is trying to maximize should also be an area of scrutiny, speculating that Microsoft's Tay bot might have behaved the way it did because it was seeking to maximize engagement. She also stated that, similar to calls for AI to be regulated as a utility, that training datasets are the real utility: standardized and easily accessible to anyone.
Bhardwaj concluded his questions by asking each panelist for a final statement about where AI will be going in the next five years. Sosak predicted that AI will be integrated into everyday life as a virtual assistant. Malhotra looked forward to AI efforts in medicine, saying that AI will help produce new treatments for many diseases. She also predicted that advances in natural language processing (NLP) will give AI human-level conversational abilities. Bhardwaj agreed, noting the use of AI in fighting COVID-19. Lila expressed a concern that regulation and governance of AI would not happen until someone was harmed by an AI system.