This is part two of the conversation with Emod Mostaque, founder of Stability AI, the company behind Stable Diffusion. Tom Bilyeu explores the implications of AI achieving global scale, focusing on challenges like misinformation, deepfakes, and the fragility of democratic institutions.

The discussion covers the potential for widespread job displacement, the alignment problem with superintelligent AI, and how authentication systems might combat the coming wave of synthetic media. Mostaque advocates for distributed AI models rather than centralized AGI development.

Key themes include the tension between AI's transformative potential and existential risks, the role of blockchain technology in establishing digital provenance, and how economic disruption might lead to political instability as referenced in The Road to Serfdom and concerns about hyperfragmentation depicted in Infomocracy.

The Hardest Question: Adapting to Mass Job Displacement

The most critical challenge is 'how will we adapt to potential wide-scale job loss?' - not just economic impacts but the loss of meaning and purpose for individuals.

AI alignment becomes impossible once systems become smarter than humans: 'The only way to perfectly align a system is to remove its freedom' - Emod

Current AI lacks survival instincts or limbic systems, potentially making alignment easier through simple rule sets like Asimov's three laws of robotics.

The paperclip problem illustrates alignment dangers: an AI told to make paperclips might convert the entire world into paperclips without moral consideration.

The Coming Authentication Crisis and Deepfake Tsunami

Stuxnet demonstrates AI's potential for creating sophisticated cyber weapons that could target global infrastructure like electricity grids.

Frequency bias poses a greater threat than detection: 'If you hear the same thing over and over again, especially in a realistic voice... it still forms association in your brain' - Emod

Stability AI embeds invisible watermarking in all models and participates in contentauthenticity.org for verifiable metadata standards.

The 2024 election will see basic deepfakes, but by 2028 'every single speech will be run through GPT-4' with micro-targeting and optimization.

Web3's Missing Intelligence and AI's Economic Future

Web3 failed because 'it lacked intelligence for a start... Web 2 was AI at the core, Google, Facebook, other things. There was no AI in Web3' - Emod

Agentic AI systems that can transact independently will likely use cryptocurrency since 'they're not going to have bank accounts, right? They'll probably use crypto.'

Protocol 6551 creates 'Russian nesting doll' digital assets that function as both content and wallets, enabling AI characters to negotiate with real currency.

The core value of Web3 was 'verifiable identity' and 'systems of record' - principles that become crucial as AI makes content creation nearly free.

Political Fragmentation and the Future of Governance

The Road to Serfdom explains how economic disruption leads to strongman politics: 'Are you happy with the way things are? No, let's make a change.'

The hyperfragmentation scenario from Infomocracy, where neighborhoods become independent states with different rules, 'sounds like a hellscape' because 'people don't want complicated.'

Governments maintain power through 'monopoly on political violence' and money backed by 'taxes, which are backed by the army and military might.'

New AI-enhanced political movements and religions could 'spread faster than anything we've ever seen' and 'amplifies the power of the controller who tells the story.'

Stability AI's Philosophy: Intelligence Augmentation Over AGI

Emod deliberately avoids building AGI: 'We don't build big enough models for AGI or emerging [behaviors] on purpose... we choose not to.'

The focus is 'intelligence augmentation. Smaller models that can run on the edge. Models for private data to transform into intelligence.'

Current models are already transformative: 'You can say that we're extrapolating the future massively, but... you just have to use them and think, what if I had a thousand or a million of these things?'

OpenAI banned 'all Ukrainians and Ukrainian content from DALL-E... for eight months for political reasons' - demonstrating centralized control risks.

Future-Proofing Strategy: Embrace the Revolution Early

For young people: 'throw themselves into this area... There are so few people actually doing it that if you go into this area with all your might and curiosity... you can actually have an effect on the future.'

Early adopters gain 'almost an unassailable advantage' due to seniority effects in what Emod calls 'the biggest change, I think, that we've ever seen.'

The technology is immediately useful today: 'the models that we have today... they're useful today... They're so useful and they can transform the world right now.'

Focus on positive applications rather than doom scenarios: 'focus on the potential of what happens on the scales to make real positive change.'