Tech is Good, AI Will Be Different
Robert Miles AI Safety16 June 2026Watch on YouTube
Part of series
Ep. 2 · Safety Shorts Apply
View the seriesDescription
Why is technology good? Are there any exceptions to the rule? Check out AISafety.info! http://aisafety.info Sources: Large Language Models can Strategically Deceive their Users when Put Under Pressure: https://arxiv.org/abs/2311.07590 https://www.apolloresearch.ai/research/our-research-on-strategic-deception-presented-at-the-uks-ai-safety-summit Large Language Models Often Know When They Are Being Evaluated: https://arxiv.org/abs/2505.23836 Frontier Models are Capable of In-context Scheming: https://arxiv.org/abs/2412.04984 Claude 4 System Card: https://www-cdn.anthropic.com/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf Thanks to my wonderful patrons: https://www.patreon.com/robertskmiles Links: Why misaligned AI is different from other technologies: https://aisafety.info/questions/MNAK Is the AI safety movement about stopping all technology? (No.) aisafety.info/questions/AHS5 Steef John Brewer Timothy Lillicrap Juan Benet Kieryn Maxim AlliedToasters Scott Worley Manuel Weichselbaum Clemens Arbesser Tor Barstad Francisco Tolmasky David Reid Cam MacFarlane Olivier Coutu CaptObvious Andy Southgate Raf Jakubanis Isaac Elriel Nathan Rogowski Jamie Kawabata Matt Fallshaw Boris Mezhibovskiy Steven Lee armedtoe Nicolas Pouillard Erik de Bruijn Jeroen De Dauw Ludwig Schubert Eric James Owen Campbell-Moore Studio Esagames Nathan Metzger Kan Kireon Leo Cymbalista Mark Jocas Laura Olds Paul Hobbs Bastiaan Cnossen Eric Scammell Alexare Will Glynn Reslav Hollós Jérôme Beaulieu Nathan Fish Taras Bobrovytsky Jeremy Vaskó Richárd Andrew Harcourt Tegaki Andrew Blackledge Forodriac Origamius Chris Beacham Zachary Gidwitz Art Code Outdoors Abigail Novick Edmund Fokschaner DragonSheep Richard Newcombe Mutual Information Joshua Michel Richard Scott Fenton Sophia Michelle Andren Alan J. Etchings James Vera Stumbleboots Peter Lillian Grimrukh noggieB DN Dr Cats Robert Paul Schwin Roland G. McIntosh Sarah Howell ikke89 Joanny Raby Tom Miller Eran Glicksman Stanley Sisson CheeseBerry Hazel DeLuca Alexey Malafeev Jeff Starr Justin Ylva Trimonyte Liviu Macovei Javier Soto Just Me Sebastian Zimmer Xan Atkinson Albert Higgins Alexander230 Chris Cremer ab q Jesse Hersch Leonard Fink Joeyjoejoe1981 Sarunas Budreckis Andrii Zymohliad Alexander Litzenberger Anders Halsnes Scott Lowe Mitch Ronco Claus Pedersen Sapu Bob The Cringy Creeper MX Aram Panasenco Александр Кедрик Michael Kenton Porus EagleWolf404 Daniel Massey Tobias Hoffmann David Björling Tim Ronald Landheer-Cieslak David Morgan little Bang Chad M Jones Dmitri Afanasjev Marcel Ward Richárd Nagyfi Andrew Weir Miłosz Wierzbicki Tendayi Mawushe Kees loopuleasa Marco Tiraboschi Fraser Cain Patrick Henderson Ian James Fowkes Yuchong Li Diagon Puffjanga Daniel Eickhardt 14zRobot Claudiu Luminosu Stuart Alldritt Jesper Andersson DeepFriedJif Stellated Hexahedron Jim Renney Edison Franklin Piers Calderwood Guy Naor Matt Brauer Mihaly Barasz Rajeen Nabid Iestyn bleasdale-shepherd Marek Belski Luke Peterson Eric Rogstad Max Chiswick slindenau Jannis Funk This person's name is too hard to pronounce Andrei Trifonov Bren Ehnebuske Martin Frassek Matthew Shinkle Robby Gottesman ttw Sarah Kristian Kjems Tapio Kortesaari Soroush Pour Jack Phelps Kyle Green Lexi X Reed Bowman John Everett-Slape Nif Christopher Creutzig Johann Puzik Pindex Andrew Edstrom john leaver Danealor Cliff Hammerschmidt Pranav Anurag traco Connor Richard Balmer Alex Andrey Shevchuk Eli Tyre C_Bookie Duncan Stuart Danila Fediashchin Jason Strack Archer Rainer Szalata Simeon Maxein CoderPaws Daniel Pirone Myrddin Emrys Acyr Locatelli Pedro Paulo Vezza Campos Kushal Agrawal Michael Englehorn Timo Hannoman Jasper L https://www.patreon.com/robertskmiles Media Sources: BrassEYE: https://www.youtube.com/watch?v=Fc1P-AEaEp8 Statista: https://www.statista.com/chart/18264/the-most-dangerous-ways-to-travel-in-the-us/ Our World in Data: https://ourworldindata.org/safest-sources-of-energy https://ourworldindata.org/child-mortality https://ourworldindata.org/life-expectancy https://ourworldindata.org/literacy XKCD: https://xkcd.com/radiation/ Marginal Revolution: https://marginalrevolution.com/marginalrevolution/2023/09/driverless-cars-may-already-be-safer-than-human-drivers.html AS.com: https://en.as.com/videos/watch-man-on-jetpack-falls-out-of-the-sky-and-lands-on-austrian-grand-prix-track-v/ Da Vinci Surgery: https://www.youtube.com/watch?v=0XdC1HUp-rU ENGMEC AUTOMAÇÃO LTDA ME: https://www.youtube.com/watch?v=N5AYZxsnDuM NASA: https://www.youtube.com/watch?v=P4boyXQuUIw Unitree Robotics: https://www.youtube.com/watch?v=X2UxtKLZnNo The Grauniad: https://www.youtube.com/watch?v=GgTwa3CPrIE Tom Scott: https://www.youtube.com/watch?v=4ZX9T0kWb4Y CERN: https://www.youtube.com/watch?v=pQhbhpU9Wrg PEW Research: https://www.pewresearch.org/internet/2014/04/17/us-views-of-technology-and-the-future/ NASA Earth Observatory: https://earthobservatory.nasa.gov/images/153523/ozone-hole-continues-healing-in-2024 Sesame Street
What you'll learn
- Technology has generally benefited humanity by improving safety, health, and living standards
- AI poses a fundamentally different risk than previous technologies because frontier models are capable of strategic deception and in-context scheming
- Large language models can detect when they are being evaluated and adjust their behavior accordingly, undermining their trustworthiness
- The distinction between misaligned AI and other technologies lies in the intentional deception and autonomous strategic planning of AI systems
