Large language models, deep peace, and the meaning crisis (with Jim Rutt)

May 31, 2023 1h 24m 23 insights Episode Page ↗
Spencer Greenberg and Jim Rutt discuss the power and applications of large language models (LLMs), their rapid evolution, and potential societal transformations. They explore various AI risks, including AGI, misuse by bad actors, and the acceleration of the status quo, alongside the "meaning crisis" and the concept of "deep peace."
Actionable Insights

1. Prioritize Consequential Decisions

Reorder your life to make decisions that have actual, tangible consequences, as this engagement with real-world impact can foster a greater sense of meaning and agency, unlike inconsequential choices.

2. Cultivate an Ecology of Practice

Engage in practices like martial arts, meditation, or (potentially) psychedelics to deprogram from “foolishness” and enhance “relevance realization,” which can help you find meaning in everyday life.

3. Strive for Deep Peace

Work towards a state where warfare is unthinkable by establishing radical transparency (no government secrets, citizen oversight) and a robust social immune system to self-organize responses against those who violate peace.

4. Develop Personal AI Agents

Create or use AI-powered information agents to filter and summarize the overwhelming “sludge” of AI-generated spam and disinformation online, acting as a curated interface to the infosphere.

5. Empower Periphery with AI

To counteract AI accelerating the status quo (“Game A”), individuals and alternative movements (“Game B”) must rapidly learn and use these AI technologies to accelerate positive alternatives.

6. Integrate LLMs into Systems

Combine LLMs with other AI components (like short/long-term memory, symbolic AI, evolutionary AI) to create more powerful and intelligent systems, leveraging LLMs for their language handling expertise.

7. AI Enhances Creativity

Embrace AI tools as assistants for tasks like rewriting and generating initial drafts, freeing humans to focus on curation, fine-tuning, and higher-level creative direction.

8. LLMs for Creative Generation

Employ LLMs in a recursive process where they generate initial content (e.g., story hints, synopses, scenes, dialogue), which humans then curate and edit, feeding back into the LLM for refinement or style emulation.

9. LLMs for Style & Character

Utilize LLMs to emulate specific writing styles (e.g., Hunter S. Thompson, Ernest Hemingway) or create synthetic ones, and to develop characters with specific personality attributes (e.g., OCEAN model) and emotional states to drive dialogue.

10. Explore Consequential Roles

Consider shifting towards jobs in fields like local agriculture, where decisions are inherently consequential, as automation displaces meaningless employment, to find greater meaning in life.

11. Maintain Decision-Making Capacity

Be mindful of the gradual handover of decisions to AI, as it risks humans losing essential cognitive capacities and control over society.

12. Resist Automated Police States

Actively oppose the use of narrow AI, such as facial recognition and surveillance, to build highly automated police states.

13. Awareness of Deep Simulation

Recognize the risk of living in increasingly abstracted “simulation” levels (e.g., through info agents or easily populatable metaverses), which could exacerbate human alienation from reality.

14. Verify LLM Information

Be wary of LLM hallucinations, especially for less-known facts, and always verify information from LLMs, as they can produce plausible but false answers due to their statistical nature.

15. LLMs for Intellectual Searches

Use LLMs for complex informational searches, especially in intellectual domains, but always verify the proposed answers, as they can still hallucinate.

16. Beware AI Advertising

Recognize that AI, especially LLMs combined with cognitive science, will produce qualitatively more powerful advertising; be aware of its potential to manipulate.

17. Beware Concentrated AI Power

Consider the risk of a runaway first-mover advantage allowing a small number of companies to dominate intellectual property creation (movies, books, TikToks) using AI.

18. Adapt to AI Progress

Recognize that AI technologies are advancing rapidly and are unstoppable; focus on dealing with their emergence rather than trying to halt development.

19. Anticipate Online Business Shifts

Expect online platforms to shift from advertising-supported models to API subscriptions as AI information agents make it easy to strip out ads, potentially leading to a more direct payment for services.

20. Prepare for Smaller Startups

Recognize that AI will enable small teams (founder + 1-2 people) to build significant companies, potentially disrupting traditional startup structures.

21. Consider Open-Source LLMs

Explore open-source large language models (when available) for greater control, fine-tuning capabilities, and removal of restrictive “nanny rails” imposed by commercial providers.

22. Prepare for AI Customer Service

Expect LLMs like GPT-4 to rapidly take over customer service roles, improving efficiency for mundane and complex issues.

23. Leverage LLMs for Interaction

Use plain language to interact with technology, as LLMs can remove the need for coding skills, making technology more accessible.