All Segments
AGI by 2028 - Shane Legg (DeepMind Founder)
3 segments available
Full Episode: https://youtu.be/Kc1atfJkiJU Apple Podcasts: https://podcasts.apple.com/us/podcast/shane-legg-deepmind-founder-2028-agi-new-architectures/id1516093381?i=1000632720307 Spotify: https://open.spotify.com/episode/0Ru2CtaJqsQ5mpA5dqHWAK?si=4AsglwIZQpqht7p9Wpc_CA Transcript: https://www.dwarkeshpatel.com/p/shane-legg Follow me on Twitter: https://twitter.com/dwarkesh_sp
Segments Timeline
The 2009 Prediction
Shane Legg reflects on his 2009 blog post predicting human-level AI by 2025, with an expected value of 2028. He discusses the context of this prediction, emphasizing the exponential growth of computational power and data, which he believed would drive advancements in AI. This segment highlights the foresight Legg had before deep learning became mainstream.
"speaking of the early years it's really interesting that in um 2009 you had a blog post where you say my modal expectation of when we get human level AI is 2025 expected value is 2028 and this is befo..."
Exponential Growth and Scalable Algorithms
In this segment, Shane Legg elaborates on two key beliefs that shaped his predictions: the exponential growth of computational power and data. He explains how these factors create a demand for scalable algorithms, which in turn leads to a positive feedback loop that accelerates AI development. This insight is crucial for understanding the dynamics of AI progress.
"exponentially for at least a few decades and that the quantity of data in the world would grow exponentially for a few decades and when you have exponentially increasing quantities of computation and ..."
Unlocking AGI in the 2020s
Shane Legg discusses the potential for training AI models on data far exceeding human experience within a lifetime. He believes this capability marks a significant step towards achieving Artificial General Intelligence (AGI) by 2028. Legg acknowledges the uncertainties in research timelines but maintains a 50% probability for this milestone, emphasizing the unpredictable nature of scientific progress.
"scalable algorithms were were to be discovered then during the 2020s it should be possible to start training models on significantly more data than a human would experience in a lifetime and I figured..."