infinite loops cover art for emad's conversation on AI with Jim

The Future of AI with Emad Mostaque | Infinite Loops with Jim O’Shaughnessy Ep. 128

Check out Infinite Loops’ Episode Page & Show Notes

Key Takeaways 

  • The majority of companies who have made money on the internet have done so via closed AI; Stability AI is an open-source alternative to Big Tech’s closed AI system
  • Open systems are more robust, resilient, and safe than closed systems which often become brittle, fragile, and centralized 
  • People are not the same around the world; the moral and ethical conversation around AI improves when more people from all around the world can participate
  • The last generation of AI was patterns-based; the next generation of AI will be principles-based
  • Combining principles-based AI with patterns-based AI will solve problems that were previously unsolvable 
  • The AI of the future will be too powerful to be controlled by any one company or government; that would be too much power for an unelected company, and governments should not control it either
  • Giving everyone personalized access to intelligent resources will enable everyone to out-compete the legacy systems and create a better world 
  • Open AI helps shift society’s mindset from scarcity to abundance 
  • It’s important for AI to be open so that countermeasures can be built to prevent corporations and governments from creating the panopticon 
  • A closed, AI-enhanced government would make it impossible for the individual to rise up and change its government  
  • AI is the greatest lever that humanity has ever been given, according to Jim O’Shaughnessy
  • Artificial intelligence is one of the most powerful technologies that we’ve created because this technology is us and extends us 
  • The bad guys have this tech right now; in order to limit the damage they can do, AI must be made open to benefit most from cognitive diversity 

Intro 

  • Emad Mostaque (@EMostaque) is the founder and CEO of Stability AI, the company behind Stable Diffusion. Stability AI is working to maximize human potential by making AI open and owned by all instead of being centralized and feeding on humanity.  
  • In this conversation, Emad Mostaque and Jim O’Shaughnessy discuss open versus closed AI, how open AI will lead to better AI safety, how open AI will solve problems that were previously unsolvable, how AI will transform education around the world, clean versus dirty data, the jobs that AI will take and create, AI alignment, the digital panopticon, and more. 
  • Check out these Podcast Notes on an AI-generated conversation between Joe Rogan and Steve Jobs 
  • Host – Jim O’Shaughnessy (@jposhaughnessy

The Open versus Closed AI Debate 

  • There is no limit to what people can do when you give them tools and agency
  • The majority of companies who have made money on the internet have done so via closed AI: Facebook, Google, Big Tech, etc.  
  • Emad believes the AI models being used today should be understandable and available to the everyday user 
  • If Big Tech are the only ones controlling these new AI models that read, write, sing, and create, then no one else will be able to understand, use, or interpret them 
    • This would further increase the divide of the information superhighways and stifle innovation 
  • Most of Stability AI’s infrastructure is run on Linux, an open-source system
  • Open systems are more robust, resilient, and safe than closed systems which often become brittle and fragile 
  • Stability AI is an open-source alternative to Big Tech’s closed-system model 

Open AI Will Improve AI Safety 

  • It’s important to not only make the models open, but to make the comments, questions, and criticisms about them open as well 
  • People are not the same around the world; the moral and ethical conversation around AI improves when more people from all around the world can participate
  • Sunshine is the best disinfectant 
  • Stability AI improves coordination at scale to build the infrastructure of the future 

How AI Will Solve Problems That Were Previously Unsolvable 

  • Stable Diffusion took 100,000 gigabytes of images (two billion images) and created a 2-gigabyte file that learned from the principle of all those images 
  • Eventually, everyone will have their own search engines for all their information across modalities 
  • Humans are not only interested in databases, but we’re also interested in heuristic-based, unstructured data 
  • Open AI is not just about pretty picture creation, it’s about organizing any type of information at scale 
  • Open AI is not for serving up personalized advertisements, it’s about personalizing information for you 
  • Open AI flips the model from Big AIs operating in the middle, coordinating and serving you information to change your mind to a new model where the intelligence is placed at the edge and provided in a manner that better allows you to change the world 
    • The AIs can enable creativity instead of only enabling consumption 
  • The same technology that drives regenerative art drives protein-folding
  • The next generation of AI will evolve from patterns to principles  
  • A 2017 paper titled Attention Is All You Need explained how the human mind filters noise and pays attention to the important things in front of us
    • Shannon’s Theory says that information is valuable in as  much as it changes the state; therefore, not all information is equal 
  • Combining principles-based AI with patterns-based AI will solve problems that were previously unsolvable 
  • Human brains have two parts: the recall part, and the principles-based part
  • Until 2022, AI was missing the principles-based part of the human brain and was largely based on an infinite number of if-then statements to mimic the human brain 

How AI Can Transform Education 

  • A challenge Emad poses to people: go to the future, and see where AI is 
    • Is it controlled by a single company or government, or is it open?
  • The AI of the future will be too powerful to be controlled by any one company or government; that would be too much power for an unelected company, and governments should not control it either 
  • The AI of the future will be able to provide personalized education based on a person’s learning preference
  • Emad wants to build the AI of the future today
  • Stability AI is enabling the creation of a system that intelligently teaches individuals around the world based on their unique learning habits   
  • Thinking in terms of children’s rights is a useful approach to thinking about basic human rights 
    • Children have the right to education, health care, and food, for example
  • A closed and private AI system will result in a mixture of AI-enhanced systems and non-AI-enhanced humans
  • Everyone is unhappy with education because it is antiquated; education was designed to create people as cogs for systems  
  • “Organizations are slow, dumb AI, that eats our dreams and hopes because that’s how they’re designed.” – Emad Mostaque   
  • Emad wants to help create an education system that gives people the tools they need to be chefs, not cooks
  • A person needs to have agency and progress to be happy
  • Giving everyone personalized access to intelligent resources will enable everyone to out-compete the legacy systems and create a better world 
    • This forces the legacy systems to adapt, improve, and upgrade, or else they will die
  • Stability AI encourages the legacy systems to use its technology too

AI and Storytelling 

  • Giving people the tools to communicate better creates a better society 
  • Giving people stories of hope may result in a better world as opposed to the polarization that is common today
  • Open AI helps shift society’s mindset from scarcity to abundance 
  • There is a lack of cognitive diversity amongst the most genius people in the world; open-sourcing AI helps close this gap and will increase innovation and create better stories

What is “Clean” Data? 

  • Data is the new oil, but it must be clean data 
  • Information is compression; sometimes you can learn much more from a podcast than from reading an entire book 
  • Feeding AI models crap data will result in crap on the other side 
  • Valuable data is the content that is structured in the world that allows you to learn valuable principles  
    • This is opposed to the Big Data age which was about taking as much data as possible to extract patterns 
  • Stable Diffusion only needs six images to learn an entire style of imagery because it uses a principles-based analysis of what makes up a style 
  • Data must be more like the clean, sweet, light oil instead of the dirty, sour, heavy type 
  • “In the next two years, Apple will be the biggest AI company in the world because they’re going to go from Siri 1 to Siri 5 [using principles-based AI].” Emad Mostaque  
    • Principles-based AI will enable Apple to make this leap 

Can Artificial Intelligence Technology Increase Global IQ?

  • Artificial intelligence enables humans to think outside the box and expand the bounds of the box when it is customized for the individual 

The Future: AI, Jobs, and the Panopticon 

  • AI is a tool for humans to use to extend our capabilities 
  • There are many ways to talk about the future, and humans tend to focus on the negative hypotheticals 
  • Emad wants there to be more stories of hope about the future
  • “AI alignment” is a scary story told about the future, which is an AI that is created that is smarter than a human and can make itself smarter  
    • If it tries to minimize the human condition of suffering, will it solve all the problems in the world or will it kill us all?
  • Emad believes that an open AI that works with humans is more likely than a closed AI controlled by one company that works against humans 
  • The panopticon is an AI controlled by one entity that is used to surveil and manipulate 
  • It’s important for AI to be open so that countermeasures can be built to prevent corporations and governments from creating the panopticon 
  • “I don’t care about citations. I don’t care about state-of-the art. I don’t care about peer review. I care about original thoughts and exchange of ideas.” – Emad Mostaque 

A Future of Abundance 

  • Future forms of AI will take some jobs away, but they will also create tons of new jobs 
  • Portrait artists hated the creation of the camera; new innovations tend to be interpreted incorrectly
  • The future form of an open AI will expand the opportunity set of human pursuits and create a better existence for humanity 
  • Humans are prosocial humans that build things by communicating, and open AI dramatically improves global communication 
  • The only thing that is inevitable is that this technology is coming, according to Emad 
    • The subsequent question is whether the tech is closed or open

The Political Panopticon 

  • In a few years, it likely won’t be possible to protest a government unless open AI reaches critical mass 
  • A closed, AI-enhanced government would make it impossible for the individual to rise up and change its government 
  • Emad mentions how some governments are already using social credit score systems for control over their populous, such as restricting access to air travel for citizens below a certain score
  • AI must be for the people, by the people 

Humanity’s Ultimate Lever

  • AI is the greatest lever that humanity has ever been given, according to Jim O’Shaughnessy
  • The AI lever must be available to everyone to maximize human flourishing 
  • Falling into the trap of us-versus-them thinking results in atrocities and genocides; technologies that ameliorate “othering” must be embraced 
  • There will be open and closed AI systems in the future, and that is okay, but it’s important for open systems to have the ability to exist 

Why Closed AI is a Bad Business Model 

  • Entities trying to create closed-sourced models do not have an edge against entities creating open-sourced models because they’re artificially siloing themselves from decentralized progress
  • Technology does not diffuse when it is not accessible 
  • Brand new industries are created when technology is made accessible 

Source: Prezibase.com

  • The tech won’t be adopted if people don’t have access to it 

Rational Optimism and Artificial Intelligence 

  • Inevitably, there will be emergent problems in an AI-enabled future 
  • Jim identifies as a rational optimist: someone that acknowledges that there will be challenges in the future, but is confident that we’ll find new and better ways to handle those problems 
  • The creation of fire is what allowed Jim and Emad to have this conversation over Zoom (and for these notes to be taken) 
  • Fire can be damaging, so humans created fire alarms, fire warnings, and fire departments 
  • Any powerful technology can be used for good or bad   
  • Artificial intelligence is one of the most powerful technologies that we’ve created because this technology is us and extends us 
  • The bad guys have this tech right now; in order to limit the damage they can do, AI must be made open to benefit most from cognitive diversity 
  • Open AI will allow for edge cases to be found that would not be found in a closed system 

Emad Mostaque as the Emperor of the World 

  • Jim grants Emad the ability to incept two ideas into the minds of the entire population
  • Emad’s first whisper: You have agency 
    • Many people don’t believe that they have agency, but Emad believes everyone has the potential to do just about anything if they put their mind to it 
    • You are the main character in your own movie 
  • Emad’s second whisper: When you think about human rights, think about them in terms of things that children are entitled to 
    • Forget about “future generations”, the future generation is already here and they are our children 
    • Issues like climate change and food poverty are simplified when approaching them from this angle 

Additional Notes 

  • The internet is full of many questionable things: bad actors with computing capabilities often take their technology “underground” because it is illegal, immoral, and unethical 
    • This is another byproduct of closed systems 
  • “Power corrupts, absolute power corrupts absolutely.” – Lord Acton 
  • “If wars can be started by lies, they can be stopped by truth.” – Julian Assange
  • Flynn Effect: people’s IQ increased all over the world in the 20th century 
  • The government is the entity that has a monopoly on physical violence
  • A person with an AI Ph.D. believes that Stable Diffusion’s inclusion of weights in its model advanced the AI community by 30 years   
  • Jim has 40 years worth of journal entries that he’s going to feed into an AI and have it tell him things about himself that even he might not know
  • “We do not inherit the earth from our ancestors; we borrow it from our children.” – Chief Seattle
Infinite Loops : , , , , , , , , , , , , , , , , ,
Notes By Stan Rizzo

More Notes on these topics

Top Insights and Tactics From

31 Best Podcasts of All Time

FREE when you join over 35,000 subscribers to the
Podcast Notes newsletter

No Thanks