Forecasting the things that matter (with Peter Wildeford)

Oct 21, 2022 Episode Page ↗
Overview

Spencer Greenberg and guest Yosha Bach explore intelligence as model-making, the nature of sentience, and how AI challenges our understanding of human minds. They delve into philosophical concepts like computational semantics and the nature of existence, offering insights into personal growth and intergroup relations.

At a Glance
13 Insights
1h 32m Duration
13 Topics
7 Concepts

Deep Dive Analysis

Defining Intelligence: Models, Control, and Generalization

Intelligence, IQ, and the Nature of Pattern Recognition

Human Causal Inference vs. AI Statistical Correlations

Embodiment, Symbol Grounding, and Meaning from Text

GPT-3 in a Robot Body: Sentience and Online Learning

Westworld Analogy: Human Identity, Freedom, and Desires

Separating Pain from Suffering: Insights from Eastern Philosophy

Methods for Self-Regulation and Understanding Values

Defining Love: Shared Sacredness and Next-Level Agency

AI as a Philosophical Project: Bridging Math and Philosophy

Gödel's Incompleteness, Computability, and Truth

The Nature of Reality: Computable Universe and Existence

Limitations of Deep Learning and the Third Wave of AI

Intelligence

Intelligence is defined as the ability to make models, typically in service of control. This allows a system to become an agent by predicting how its actions will lead to different future outcomes and preferring some over others, distinct from rationality, smartness, or wisdom.

Sentience

Sentience is the discovery of one's own nature as an agent in the world and the relationships one has to that world. It occurs when an intelligent system develops a model where it discovers itself in relation to its environment.

Vectors of Intelligence

Instead of a single linear scale, intelligence can be described across multiple dimensions, such as the capacity for autonomy, control, perception, reasoning, language learning, embodiment, collaboration, and knowledge representation. This allows for a richer comparison of different intelligent systems like humans, animals, and AI.

Causal Structure Discovery

This refers to the human ability to uncover the underlying systems that produce observed patterns, rather than merely recognizing statistical correlations. It is crucial for understanding how the world works and often relies on tracking changes and the flow of information over time.

Computational Semantics

A modern philosophical view where truth is understood as a stateful notion, assigned by executing an algorithm, rather than a timeless, platonic concept. This perspective implies that truth is tied to the procedure by which it is acquired and can change based on the sequence of states.

Shared Sacredness (Love)

This describes a form of love as the discovery of shared purposes that transcend individual ego, for which one is willing to sacrifice. It enables non-transactional interactions and the creation of 'next-level agency,' akin to how individual cells make sacrifices for the organism's existence.

Differentiable Programming

The underlying principle of deep learning, where programs (like neural networks) are written to describe a somewhat continuous state space. This allows solutions to problems to be found by following gradients, meaning small, predictable changes can be made to the program's parameters to move towards a desired output.

?
What is intelligence?

Intelligence is defined as the ability to make models, typically in service of control, allowing a system to predict future outcomes of its actions and choose preferred branches.

?
How does human intelligence differ from current AI models like GPT-3?

Human intelligence involves algorithmic modeling, causal inference, and embodiment, while current AI models primarily focus on statistical correlations, often lacking true embodiment, online learning, or the ability to develop algorithms to model other algorithms.

?
What is the relationship between pain and suffering?

Pain is a sensation, whereas suffering arises from a mismatch between what one tries to regulate and what can actually be regulated, often linked to narratives or stories one tells oneself about the pain or situation.

?
How can one overcome emotional suffering, such as heartbreak?

One can overcome emotional suffering by writing out thoughts to stop circular thinking, which allows for higher-level abstraction and a deeper understanding of the situation, including the perspectives of others and one's own role, leading to disentanglement from the immediate emotional state.

?
What is the 'philosophical project' of AI?

The philosophical project of AI aims to bridge mathematics and philosophy by building machine minds that can perform philosophy with higher acuity and detail than humans, ultimately seeking to understand what minds are as a mathematical model that can be automatically executed.

?
How does Gödel's incompleteness theorem relate to the nature of truth?

Gödel's proof implies that truth is not a stable, platonic concept but rather a stateful notion, a predicate assigned by executing an algorithm. Languages that contain infinities as if they existed become self-contradictory, suggesting that only computable results have true values.

?
What does it mean for something to 'exist' from a computational perspective?

From a computational perspective, for something to exist means it needs to be implemented. Its existence is tied to the degree of its implementation and the ability to describe it within a consistent language, often implying a causal structure that gives rise to regular observations.

?
What are the limitations of current deep learning models?

Current deep learning models are not sample efficient, prone to overfitting (leading to adversarial examples), struggle with efficiently learning causal structures or programs, and rely on hardware 'hacks' rather than optimal architectures for intelligence.

?
What is the 'third wave' of AI expected to achieve?

The third wave of AI is expected to produce systems that can extend themselves into the world, understand and create languages, integrate deeply with people, use more universal representations, and perform program synthesis as effectively as they follow gradients.

1. Challenge Fixed Identity

Question your fixed identity (e.g., gender, personal story) to overcome self-imposed limitations and understand your full potential, as a rigid identity can prevent you from seeing other possibilities for yourself.

2. Alleviate Suffering by Regulating

Reduce suffering by improving your models of how the world works and your place within it, as suffering often arises from attempting to regulate things that are beyond your control.

3. Separate Pain from Suffering

Practice viewing pain as a neutral sensation rather than inherently ‘bad’ to create a distinction between the physical experience and the emotional suffering, fostering greater acceptance.

4. Process Emotional Pain by Writing

When experiencing intense emotional pain, such as heartbreak, sit down and write out your thoughts to prevent repetitive rumination and access higher-level understanding and more effective coping behaviors.

5. Resolve Disagreements by Elevating Perspective

To resolve disagreements, ascend at least two levels of abstraction to understand the underlying values and their construction in both yourself and others, which facilitates deeper understanding and negotiation.

6. Cultivate Multi-Generational Perspective

Perceive yourself as part of a multi-generational entity, such as a family line or a broader societal project, to foster a deeper sense of purpose and love that extends beyond individual desires.

7. Seek Good in All Worldviews

Actively look for positive aspects and valid ideas in all worldviews, including those of ‘out-groups,’ to gain a more accurate understanding of the world and avoid the belief that only your in-group holds the correct answers.

8. Develop Self-Understanding for Growth

Reverse engineer your own mind to build deeper structures of understanding, which enables personal growth and more reasonable interactions, such as evolving from infatuation to love based on shared purpose.

9. Test Child’s Programming Aptitude

For children, giving them the task to write programs from scratch, even without prior learning, can serve as a valuable predictor for later cognitive performance in life.

10. Adopt Model-Making Intelligence View

Consider intelligence primarily as the ability to make models, distinguishing it from rationality, smartness, or wisdom (the ability to pick the right goals), to gain a clearer understanding of cognitive abilities.

11. Understand Information as Change

Recognize that the meaning of information is fundamentally its relationship to change and other information, as the brain primarily registers and models changes to construct an understanding of stability.

12. Persist in Learning from Text

Be aware that deriving meaning and ‘signal’ from purely textual symbols is possible but often requires significantly more time and exposure compared to learning through embodied interaction with the world.

13. Embrace Computational View of Reality

Adopt a computational perspective where existence implies implementation and truth is a stateful, algorithmic process, rather than a timeless, platonic concept, to develop internally consistent languages for describing reality.

Personally, I think of intelligence as the ability to make models, and it's distinct from being rational, right? A lot of irrational people are highly intelligent and vice versa.

Yosha Bach

Sentience is not the same thing as consciousness. It's basically just a discovery of your own nature as an agent in the world and the relationships that you have to the world.

Yosha Bach

The true message of Westworld is not a story about robots. It's a story about ourselves. Because we are like these robots. We are not actually human beings. We are not actually hairless monkeys. What I am is a mind. I am a side effect of the regulation needs of a monkey. I just happen to run on a monkey's brain. I am not that monkey. I can be whatever I want to be if I stop believing in the stories of the monkey and the desires of the monkey.

Yosha Bach

Suffering is not the result of the universe doing something to you as an agent. The universe that you experience is not the physical universe. It's not some weird quantum pattern or something. It's a world full of meanings, of desires, of stories that we have already chosen to adhere to before we got to the discovery of our own self.

Yosha Bach

The benefits of philosophy tend to be tiny. And even though I think it's the most important philosophical project of all, philosophy, for the most part, is not that important to humanity. Most of the big questions are no longer philosophical questions.

Yosha Bach

Pi is not a value. Pi is just a function from the computational perspective, which means you can plug your machine that computes digits of pi into an energy source, into your local sun. And when the sun burns out, this is your last digit.

Yosha Bach

I don't think that you are, if you do philosophy or if you do modeling as a mathematician, you are entitled to arbitrary ambiguity. I think that's a cop-out.

Yosha Bach