The Ethical Dilemma of Ensuring Unbiased AI DevelopmentRichard Foster-Fletcher on a Conscious and Inclusive Approach to AI Development
Earlier in October, Demis Hassabis, CEO of DeepMind - an AI research laboratory and Google subsidiary - urged for AI risks to be addressed with the same urgency as climate change. The current discourse on the subject is not sufficiently wide-ranging, inclusive, probing and uncomfortable, said Richard Foster-Fletcher, executive chair, Morality and Knowledge in Artificial Intelligence (MKAI).
The rapid advancement of AI brings forth myriad challenges, including the complexities of AI ethics. The challenge lies in the inherent nature of AI to simplify and categorize, potentially reinforcing existing biases through the data it was trained on.
The platform [ChatGPT] is predominantly trained on English language content from the internet, which largely originates from the U.S. and is developed by a male-dominated workforce, Foster-Fletcher said. It inherently reflects a Western-centric perspective, which raises concerns about its appropriateness and relevance for international users, he said.
To avoid reinforcing these biases and ensure AI does not lead to discriminatory or unfair outcomes, he emphasized the need to engage in collaborative efforts among governments, organizations and diverse communities.
Foster-Fletcher dismissed the recent pledges made by big tech corporations to halt AI development as mere posturing, citing instances of contradictory lobbying efforts by leaders of prominent AI organizations.
In this video interview with ISMG, Foster-Fletcher discussed:
- The big tech challenge in navigating the ethical dilemmas in AI development;
- Assessing the merits of a temporary halt to AI advancement;
- The risk of potential biases in generative AI models.
Foster-Fletcher is one of the leading advocates of AI ethics globally. At MKAI.org, he empowers organizations to harness the power of AI responsibly while also providing guidance on potential pitfalls such as bias, exploitation and privacy infringement in gen AI applications.