#55: Skype's Cofounder on the Threat of AI – Jaan Tallinn

How many people do you know who have created a verb? We ‘google’ things, we ‘uber’ to places, and thanks to this episode’s guest, we also ‘skype’ each other.

Jaan Tallinn is the cofounder of Skype and Kazaa. But his entrepreneurial success in connecting the world pales into comparison next to his new goal: preventing the possible extinction of the human race due to unsafe Artificial Intelligence.

To that end, Jaan has cofounded Cambridge University’s Centre for the Study of Existential Risk, MIT’s Future of Life Institute, and is an investor in and director of DeepMind, among other AI startups. He has a background in theoretical physics.

Joe and Jaan discuss lessons and stories from his Skype days, why AI is an existential risk, and how we might solve the control problem.

Show notes

Selected links

• Other AI episodes: #51 and #45
• 80,000 Hours
• Skype and Kazaa
• Metcalfe’s law: the value of the network equals the number of users squared
• Eliezer Yudkowsky: Website | Twitter
• LA-602
• The ‘Concrete Problems in AI Safety’ paper
• Paul Graham’s maker vs manager schedule

Topics discussed

• What was it like growing up in Estonia under Soviet occupation? [6:15]
• Jaan’s interest in computer programming developed before high school. [8:47]
• Jaan meets a future Skype cofounder at high school. [10:48]
• Jaan’s theoretical physics background, and his thesis on interstellar travel. [12:19]
• What university degree would Jaan recommend to someone graduating high school today? [15:34]
• The story of founding Skype. [17:42]
• How did Skype scale so quickly? [26:57]
• How did Jaan’s life change after selling Skype to eBay for $2.5 Billion? [28:15]
• The three impacts Skype has had on Estonia’s startup ecosystem. [31:36]
• Why did Jaan decide AI research was the best use of his time and money after Skype? [33:35]
• Jaan discovers the writings of Eliezer Yudkowsky. [35:59]
• What are the two categories of existential risks? [39:22]
• Why do future generations matter here? [42:23]
• Why is AI an existential risk? [44:25]
• What is ‘corrigibility’ and is it the best approach to solving the control problem? [46:12]
• What do we need to do more of or better to prevent dangerous AI? [50:19]
• AI’s Nash Equilibrium. [52:10]
• In Jaan’s experience, what does the public most commonly misunderstand about AI? [57:21]
• What is Jaan working on at the moment? [1:00:24]
• Does Jaan use any heuristics for angel investing decisions? [1:03:34]
• How Jaan uses the maker/manager schedule. [1:04:56]
• The toughest moments of Jaan’s career and what he told himself to stay optimistic. [1:08:06]
• How has Jaan’s life turned out differently to what he expected it to? [1:12:19]
• Why didn’t Jaan just retire? [1:13:13]