People are worried that AGI is maybe next week or just around the corner...
ฝัง
- เผยแพร่เมื่อ 18 ต.ค. 2024
- Excerpt from: #-1 - AGI (with David Deutsch)
Link to Full Episode: reasonisfun.po...
Creator: Lulie Tanett - Reason Is Fun Podcast
Date of Publication: Saturday, Apr 08, 2023
Big shoutout to Lulie Tanett, the host of the Reason Is Fun Podcast, for the awesome content in this clip. Keep rocking it!
I listen to anything by David to ground my brain & mood… Bravo!
I enjoyed reading David Deutsch's books and I'm so happy to come across this whole series of videos. I also like the style of this interviewer (and others), always ready to challenge David Deutsch with even the simplest question or request for clarification, which is the real hallmark of the scientific temperament.
I have been listening to Reason Is Fun for a few months now, and I must say it is one of the most engaging and enlightening podcasts I have ever heard. Lulie Tanett have fascinating conversations about topics ranging from emotions, AGI, reason and fun, to epistemology, morality and enjoyment. They explore how these fields are all deeply interconnected and compatible, and how rationality and fun are not mutually exclusive, but rather mutually reinforcing. They also challenge common misconceptions and myths about science, philosophy and human nature, and offer fresh and original perspectives that are both rigorous and accessible. Reason Is Fun is a podcast that will make you think, laugh, and learn something new every episode. I highly recommend it to anyone who is curious about the world and themselves.
Drawing an analogy between the complementary nature of Popperian and Bayesian epistemologies in the development of AGI and the computer science concepts of depth-first search (DFS) and breadth-first search (BFS) is insightful. Both sets of approaches, in their respective fields, illustrate different strategies for exploring and solving problems, each with its unique advantages and scenarios where they are most applicable.
Depth-First Search (DFS) and Popperian Epistemology
DFS explores a graph or tree structure by moving as far as possible down each branch before backtracking. This approach can be likened to the Popperian method of conjecture and refutation, where a theory (or hypothesis) is pursued deeply, testing it against as many scenarios and pieces of evidence as possible to seek falsification. Just as DFS exhaustively explores one path to its end before trying another, Popperian epistemology emphasizes the depth of critical testing for each hypothesis, seeking to disprove it with thorough examination before moving on to another theory.
This methodology ensures that if a path (or hypothesis) is viable, it is explored in its entirety, understanding its implications deeply before any limitations or errors lead to backtracking. In AGI development, this could translate to an AI system probing its strategies or beliefs deeply, testing them extensively against the environment and data to identify any faults before revising its approach.
Breadth-First Search (BFS) and Bayesian Epistemology
BFS, on the other hand, explores all the neighbors of a node before moving on to their neighbors. It effectively maps out all possible paths from the starting point in a systematic, level-by-level manner. This approach mirrors Bayesian epistemology's method of updating beliefs based on new evidence, considering all available hypotheses and incrementally adjusting their probabilities. It’s akin to exploring a wide array of possible solutions simultaneously, gradually honing in on the most likely ones as more evidence is gathered.
In the context of AGI, employing a BFS-like (Bayesian) strategy would mean an AI system could simultaneously entertain multiple hypotheses or strategies, updating its confidence in each as it acquires more data. This approach allows for a more flexible adaptation to new information, paralleling how BFS navigates a problem space by considering a broader range of possibilities at each step.
Complementarity in AGI Development
Just as DFS and BFS are complementary strategies in computer science, useful for different problems or at different times, Popperian and Bayesian epistemologies offer complementary approaches to understanding and developing AGI. Employing both approaches could yield a more robust and versatile AGI:
Popperian/DFS-like approaches could drive the system to deeply understand and critically evaluate each hypothesis or strategy, ensuring thorough exploration and testing of ideas.
Bayesian/BFS-like approaches could enable the system to remain agile, updating its beliefs and strategies in light of new data, effectively navigating through a vast space of possibilities.
Integrating these methodologies could lead to AGI systems capable of deep, critical reasoning while also being adaptable and responsive to new information. This synergistic approach not only encapsulates the essence of human-like reasoning and learning but also leverages the strengths of each strategy to address the multifaceted challenges of achieving AGI.
For the same piece of bread, how many thoughts possible?
The danger of AI is not only in the addition of the general, many of the most dangerous existing intelligences are so more due to their power (i.e agency) and the limited nature (biased, closed, fixed) of their intelligence.
David thinks that AGI evolved in humans from the meme replication mechanisms of our chimp-like ancestors (re: Richard Byrne). Does that mean that a path to AGI could be found in a small (and so evolutionarily plausible) tweak to a simulation of those mechanisms? Why not?
How much do you wanna bet that it isnt.
Science is honest or it is not honest. You Tube is emotional like religion. AGI A Godlike Intelligence.
Interviewer is annoying
Arbitrary interruptions which change the subject, or spontaneous nonsensical questions based on the last keyword she understood....
Not a fan of her tryhard girlish playfulness re: breakfast cereal
@@MediocritesTheNormal Nutricious neutrinos, also rhymes with Cheerios, word associations.. guess that was on topic just about.
Agreed.
Typical UK thinking these days i.e. none or tic tok - lol.
What is agi?
Universal creator