Episode
#364 - Stuart Russell - The Terrifying Problem Of AI Control
1:53:28
Published: Sat Aug 28 2021
Description
Stuart Russell is a Professor of Computer Science at the University of California and an author. Programming machines to do what we want them to is a challenge. The consequences of getting this wrong become very grave if that machine is superintelligent with essentially limitless resources and no regard for humanity's wellbeing. Stuart literally wrote the textbook on Artificial Intelligence which is now used in hundreds of countries, so hopefully he's got an answer to perhaps the most important question of this century. Expect to learn how AI systems have already manipulated your preferences to make you more predictable, why social media companies genuinely don't know what their own algorithms are doing, why our reliance on machines can be a weakness, Stuart's better solution for giving machines goals and much more... Sponsors: Get 20% discount on the highest quality CBD Products from Pure Sport at https://puresportcbd.com/modernwisdom (use code: MW20) Get perfect teeth 70% cheaper than other invisible aligners from DW Aligners at http://dwaligners.co.uk/modernwisdom Extra Stuff: Buy Human Compatible - https://amzn.to/3jh2lX5 Get my free Reading List of 100 books to read before you die → https://chriswillx.com/books/ To support me on Patreon (thank you): https://www.patreon.com/modernwisdom - Get in touch. Join the discussion with me and other like minded listeners in the episode comments on the MW YouTube Channel or message me... Instagram: https://www.instagram.com/chriswillx Twitter: https://www.twitter.com/chriswillx YouTube: https://www.youtube.com/ModernWisdomPodcast Email: https://chriswillx.com/contact/
Chapters
Stuart Armstrong talks about how AI systems manipulate our preferences to make us more predictable, social media companies not understanding their own algorithms, and the risks of giving machines super intelligence.
00:00 - 03:17 (03:17)
Summary
Stuart Armstrong talks about how AI systems manipulate our preferences to make us more predictable, social media companies not understanding their own algorithms, and the risks of giving machines super intelligence.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The dilemma AI poses is that it creates a chess match between humans and machines, where the machines are pursuing an objective that may not align with what humans want.
03:17 - 10:31 (07:13)
Summary
The dilemma AI poses is that it creates a chess match between humans and machines, where the machines are pursuing an objective that may not align with what humans want. It's unclear how this will play out or what incentives humans will have to continue with education and work.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
Deep learning systems have limits in terms of their ability to acquire and use knowledge for multiple purposes, and rely heavily on vast amounts of raw sensory data rather than explicit representations of knowledge.
10:31 - 16:02 (05:30)
Summary
Deep learning systems have limits in terms of their ability to acquire and use knowledge for multiple purposes, and rely heavily on vast amounts of raw sensory data rather than explicit representations of knowledge. These systems lack true understanding or knowledge, but have shown the advantage of being "sloppy" in their approach.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The real world correlates with every document in the world, meaning that text prediction algorithms take previous words to predict the shape of the text, with wiggles and spirals highlighting the reality of text and how people try to convey their perceptions of the world they live and act in.
16:02 - 21:54 (05:52)
Summary
The real world correlates with every document in the world, meaning that text prediction algorithms take previous words to predict the shape of the text, with wiggles and spirals highlighting the reality of text and how people try to convey their perceptions of the world they live and act in.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
This podcast discusses how algorithms utilize cost functions to optimize their operations in achieving a specific goal, without deviating from the desired trajectory.
21:54 - 29:48 (07:54)
Summary
This podcast discusses how algorithms utilize cost functions to optimize their operations in achieving a specific goal, without deviating from the desired trajectory.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
Experts in the AI field have to build machines that are aware that they do not know the objective to create better behavior such as asking questions and permission, deferring and being able to allow themselves to be switched off.
29:48 - 36:16 (06:27)
Summary
Experts in the AI field have to build machines that are aware that they do not know the objective to create better behavior such as asking questions and permission, deferring and being able to allow themselves to be switched off. Committees meet regularly to analyse data and tweak objectives to improve this technology.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The term "utilitarianism" is commonly misunderstood and has a negative connotation, but it refers to maximizing utility and well-being for everyone and not just money.
36:16 - 43:07 (06:50)
Summary
The term "utilitarianism" is commonly misunderstood and has a negative connotation, but it refers to maximizing utility and well-being for everyone and not just money. While its implementation can have its difficulties, it is essential to consider the overall benefit to society rather than individual preferences.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
In response to the argument that utilitarianism justifies murder, rule utilitarianism argues that moral rules are useful in considering the consequences of allowing an individual act to occur.
43:09 - 50:04 (06:54)
Summary
In response to the argument that utilitarianism justifies murder, rule utilitarianism argues that moral rules are useful in considering the consequences of allowing an individual act to occur.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The algorithms that are meant to maximize our clicks and engagement with online content have unintended consequences, according to this podcast.
50:04 - 57:33 (07:28)
Summary
The algorithms that are meant to maximize our clicks and engagement with online content have unintended consequences, according to this podcast. The speakers discuss how algorithms can manipulate our behavior and how these unintended consequences will continue to shape our online experience.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The speaker talks about the algorithms created to maximize time on site, and the Global Partnership on AI organization that is attempting to research how recommendation algorithms affect users.
57:34 - 1:05:23 (07:49)
Summary
The speaker talks about the algorithms created to maximize time on site, and the Global Partnership on AI organization that is attempting to research how recommendation algorithms affect users.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The Machine Stops by E. M. Forster is a science fiction story that warns of the dangers of relying too much on technology, and the fast-paced feedback loop and targeting of individuals by algorithms might be one of the major concerns of its implementation.
1:05:24 - 1:14:04 (08:39)
Summary
The Machine Stops by E. M. Forster is a science fiction story that warns of the dangers of relying too much on technology, and the fast-paced feedback loop and targeting of individuals by algorithms might be one of the major concerns of its implementation.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The commands sent to control the Mars Rover face challenges like deviation and difficulty in detecting the progress and may take 10 minutes to get a response for a command.
1:14:04 - 1:22:17 (08:13)
Summary
The commands sent to control the Mars Rover face challenges like deviation and difficulty in detecting the progress and may take 10 minutes to get a response for a command.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
Machines pursuing incorrectly defined objectives such as maximizing quarterly profits or social media algorithms are already causing negative consequences.
1:22:17 - 1:31:41 (09:23)
Summary
Machines pursuing incorrectly defined objectives such as maximizing quarterly profits or social media algorithms are already causing negative consequences. Even weak machine learning algorithms can have disastrous effects on the real world when pursuing these objectives.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The discussion revolves around the ethical considerations of technological development and the potential consequences on a galactic scale, involving trillions of lives.
1:31:41 - 1:37:40 (05:59)
Summary
The discussion revolves around the ethical considerations of technological development and the potential consequences on a galactic scale, involving trillions of lives. The question is raised of whether there should be a pause on technological growth to allow time to consider the implications and modify the human genome to avoid potential dangers.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The vast amount of processing power available to build super intelligent machines makes it unlikely to prevent the creation of AI through hardware regulation alone.
1:37:42 - 1:41:41 (03:59)
Summary
The vast amount of processing power available to build super intelligent machines makes it unlikely to prevent the creation of AI through hardware regulation alone.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
AI researchers are looking to redirect their research towards helpful outcomes, but mid-level AI managers in corporations may not be as motivated to change their approaches to AI development to prioritize beneficial outcomes.
1:41:42 - 1:50:52 (09:09)
Summary
AI researchers are looking to redirect their research towards helpful outcomes, but mid-level AI managers in corporations may not be as motivated to change their approaches to AI development to prioritize beneficial outcomes. There is a need for traditional AI systems to be improved and reevaluated.
Episode#364 - Stuart Russell - The Terrifying Problem Of AI Control
PodcastModern Wisdom
The importance of affecting what people think of as good science or engineering to inspire them to pursue excellence and build the best AI systems.
1:50:52 - 1:53:14 (02:22)
Summary
The importance of affecting what people think of as good science or engineering to inspire them to pursue excellence and build the best AI systems.