85. Timnit Gebru Looks at Corporate AI and Sees a Lot of Bad Science
Aug 16, 2023
auto_awesome
Dr. Timnit Gebru's pioneering work in AI research, her controversial dismissal from Google, and her critique of large language models. The podcast also explores the environmental impact and biases in AI models, the importance of understanding data and representation, and the need for inclusive machine learning systems.
Curating AI datasets to ensure diversity and avoid biases is crucial for responsible development.
Challenging deceptive practices and understanding limitations of AI systems are necessary to prevent dangerous misconceptions.
Deep dives
Dr. Timnit Gebru's Firing from Google
Dr. Timnit Gebru, an influential figure in AI, was fired by Google in 2020. She co-authored the Stochastic Parrots paper, which expressed concerns about the rush to build larger language models without adequate consideration of the dangers and limitations. The paper highlighted issues such as the environmental impact of training these models, the lack of diverse viewpoints represented in the data, and the filtering of certain content. Dr. Gebru's firing brought attention to the challenges of addressing bias and ethics in AI development within Google and the wider industry.
The Need for Curation in AI Data Collection
Dr. Gebru emphasized the importance of curating AI datasets to ensure diversity and avoid perpetuating biases. The paper discussed the downsides of laissez-faire data collection, where data is gathered from the internet without careful curation. It highlighted the misconception that large datasets automatically lead to diverse viewpoints, as marginalized groups are often underrepresented. The paper also addressed the environmental impact of collecting and using vast amounts of data, raising concerns about sustainability and environmental justice.
Notions of Emergence and General Intelligence
Dr. Gebru challenged claims that large language models are achieving general intelligence or exhibiting emergent behavior. She criticized the lack of transparency and scientific rigor in such assertions, highlighting the absence of proper testing on novel data. Furthermore, she questioned the deceptive practices of corporations, promoting exaggerated capabilities of language models, which can lead to dangerous misconceptions. Dr. Gebru emphasized the need for critical evaluation and understanding of the limitations of AI systems, advocating for responsible development and avoiding hype-driven narratives.
Importance of Inclusion in AI Development
Dr. Gebru emphasized the significance of diversifying the field of AI and ensuring inclusion in decision-making processes. She highlighted initiatives like Black in AI and the Distributed AI Research Institute as efforts to involve underrepresented communities and challenge the dominance of privileged groups. Dr. Gebru argued that diverse perspectives are crucial for creating ethical AI systems that address societal needs and avoid harmful biases. By including those affected by AI in the development process, more effective and beneficial AI tools can be built.
Timnit Gebru is not just a pioneering critic of dangerous AI datasets who calls bullshit on bad science pushed by the likes of OpenAI, or a tireless champion of racial, gender, and climate justice in computing. She’s also someone who wants to build something different. This week on Reimagining, we talk to the thrilling, funny […]
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode