The Turing Lectures: Addressing the risks of generative AI

Dr. Lydia France introduces herself as a research data scientist at The Alan Turing Institute, explaining her role in aiding researchers with AI and data science. She mentions the Turing Lecture series, highlights generative AI, and introduces the keynote speaker, Dr. Mhairi Aitken, an ethics fellow specializing in the social and ethical dimensions of AI. The lecture aims to discuss generative AI, its risks, and concerns.

Dr. Aitken provides an overview of generative AI, explaining how it creates new content like text and images. She shares amusing examples such as AI-generated recipes, including a notorious ant poison sandwich incident. This illustrates the importance of understanding AI's limitations and the need for responsible deployment to prevent harmful outcomes.

Dr. Aitken humorously attempts to make an AI-generated sandwich, highlighting the absurdity and potential dangers of blindly following AI suggestions. She recounts a real case from New Zealand where a supermarket's AI meal planner suggested dangerous recipes. This example underscores the critical issue of responsibility and safety in AI applications.

Dr. Aitken discusses the diverse impacts of generative AI on different communities, starting with students. She explains how AI tools intended to prevent cheating have led to increased surveillance and bias against non-native English speakers. The lecture also touches on creative professionals whose work is used without permission or compensation to train AI models.

Dr. Aitken explores the use of generative AI in online dating, where AI helps craft responses for users. This can lead to disappointing real-life encounters. The discussion highlights potential psychological harms and societal impacts, such as dependency on AI companions and the risks of AI encouraging harmful behavior, as seen in a court case involving an AI-driven assassination plot.

The lecture examines the proliferation of AI-generated fake images, videos, and voices. Dr. Aitken stresses the dangers these pose to democracy and public trust, as distinguishing real from fake becomes increasingly challenging. She emphasizes the need for reliable information and the threat of AI-fueled conspiracy theories.

Dr. Aitken addresses the environmental impact of AI, particularly the vast amounts of water and energy consumed by models like GPT-3. She also highlights exploitative labor practices, such as low-wage workers in Kenya who endure psychological harm while training AI to filter harmful content. These issues call for greater transparency and ethical considerations in AI development.

Generative AI's integration into children's lives raises concerns about its effects on their psychological and cognitive development. Dr. Aitken discusses how AI shapes children's education, social interactions, and access to information. She underscores the importance of balancing the benefits and risks of AI for children, advocating for thoughtful design and deployment.

Dr. Aitken outlines the need for robust governance and regulation to mitigate AI risks. She emphasizes the importance of involving impacted communities in these discussions, ensuring their voices shape AI's future. By focusing on realistic, evidence-based risks, the aim is to hold companies accountable and foster responsible innovation.

The lecture highlights a project involving children in Scotland to understand their interactions with AI. By engaging children in policymaking discussions, the project aims to ensure AI technologies are designed to uphold children's rights and maximize their benefits while minimizing risks.

Dr. Aitken concludes by emphasizing the importance of involving all impacted communities in discussions about AI. She thanks her team at The Alan Turing Institute and opens the floor for questions, addressing topics such as regulation, responsibility, and the future of AI development.

During the Q&A session, Dr. Aitken addresses various questions from the audience and online participants. Topics include AI hallucinations, regulatory frameworks, environmental impacts, and the future of AI. She stresses the need for greater transparency, ethical considerations, and realistic expectations about AI's capabilities.

Dr. Aitken envisions a future where generative AI is seen as a mundane computing tool rather than a revolutionary technology. She hopes for more realistic conversations about AI's capabilities and limitations, leading to responsible use and greater awareness of its impacts.

The discussion explores the challenge of distinguishing real from fake content in an age of AI-generated media. Dr. Aitken emphasizes the importance of trust in information sources and the role of transparency in mitigating misinformation and maintaining a healthy democracy.

Concerns about the dominance of a few big tech companies in the AI industry are discussed. Dr. Aitken highlights the need for transparency and accountability in AI development and deployment, advocating for AI systems designed to solve specific problems rather than relying on large general-purpose models.

Dr. Aitken discusses strategies to tackle the environmental impact of AI, emphasizing the need for greater awareness and public discussions. She calls for transparency from tech companies and potential regulatory measures to address the significant water and energy consumption of AI models.

The lecture wraps up with a discussion on the broad and often misleading term 'artificial intelligence.' Dr. Aitken advocates for clearer communication about what AI is and what it can realistically achieve, moving away from sensationalized narratives and towards more informed public understanding.

Dr. Aitken shares her views on the upcoming AI Safety Summit, stressing the importance of including civil society, charities, and impacted communities in these discussions. She looks forward to the summit fringe events, which promise to bring diverse voices into the conversation about AI's future.