# Spark File * Simplify down to the simplest possible situation that contains the essential kernel of the problem. Solve that problem. Then add the complications back in and see how they impact the answer to the simple problem (pg 27, Biggest Ideas in the Universe: Space, Time, Motion) * What if you founded a research lab whose them was doing good science in startup environments? * Taylor series as basis functions? - Prove local linearity and put it in obsidian. Does 2nd derivative come into play? This relates to curvature…? What geometries might not have this property? - Epsilon delta definition to obsidian [https://calcworkshop.com/limits/epsilon-delta-definition/](https://calcworkshop.com/limits/epsilon-delta-definition/). the key idea is: for _any epsilon_ we can always pick a specific delta such that f is within epsilon of L. The key here is how epsilon can be _anything_ while delta is selected after that-it can’t necessarily be anything, it is constrained logically based on epsilon - computational irreducable -& reduciability - is this really talking about constraints and structure that we can exploit? - A wave requires something to propagate through! In the ocean its water, with sound its air - Would deutsch agree with the iron rule? Reconsider his chapter on explanation. I think he disagree about Strevens view on Explanatory Ore - Strevens walks dangerously close to being an instrumentalist. See [here](https://chatgpt.com/share/0b4aca91-5dd1-4509-9b56-71df1b2b2e9d). - Blog post prompt- how would you get science off the ground? How would you motivate it? Descartes examples are good. What is science? Why would a start up need to embrace it? - Why do we fall? Gravity is a beautiful example of how science works, the history being critical as well - Again, how would deutsch respond to this book? The author seems to be saying prediction is what matters, but I am not 100% sure - For a good example of why LLMs aren’t intelligent, see [here](https://youtu.be/UakqL6Pj9xo?t=1280). Patel talks about how he believes that what an LLM is doing is intelligent when solving a math question. But in reality, it simply memorized an algorithm, based on tons of training data. Yes it was given a new input (prompt) and produce a new output. But your calculator can even do that! Can it generate a new algorithm on its own via reasoning? If it doesn’t match the pattern? No. Its not creative. - Why don’t I think that AI will be able to do scientific discovery? It has not managed to perform well in constrained reasoning tasks without a lot of sample data. That requires a different type of creativity. In the arts, creativity is required, but it is combinatorial and has reduced constraints. In the sciences there are often very challenging constraints on a problem that _matter_. You can’t simply ignore them. - Popper podcast: [https://overcast.fm/+AAmesWrRQ3I/35:10](https://overcast.fm/+AAmesWrRQ3I/35:10) - Mindscape podcast: [https://overcast.fm/+AAS_7ngaKek/1:14:40](https://overcast.fm/+AAS_7ngaKek/1:14:40) - How to make more knowledge and argument explicit, not implicit? How to make it more clear what problem you are always trying to solve? A UI like this would be incredible - Egan - If I do nearest neighbors from centroids does that partition the space into chopped up hyperplanes? I believe yes. See hexagon argument of 81 egan - Pg 82 - can you be captured by freedom? This feels like an interesting question to try and use my new deutsch logic on. Is it a paradox? Logically unsound? Self reference?  - 1:37:30 deutsch Harris (first convo - surviving) - [https://youtu.be/yf-zJf2yQrU?si=0Y71Tv8iIhXJDeXA](https://youtu.be/yf-zJf2yQrU?si=0Y71Tv8iIhXJDeXA) (26:00) - Luminous pg 133 - ask if Deutsch would say that a proof is a proof if it popped into existence randomly, but no one was there to interpret it? - If I have a system of inference (logical consistency required), and I start with axioms and a set of rules, are all inferences/consequences of these rules existing even before they are proved?  - Do consequences need to be pushed into existence? Does Dr Johnson’s criteria help us here? - Reality of abstractions touches on causes, could be useful for dust theory refutation  - [http://youtu.be/2BLo2SdmjLI](http://youtu.be/2BLo2SdmjLI) 4:30 good example of infinite regress - Is it virtual reality that enables us to interact with abstract things? Without VR would we never be able to interact with abstractions? - Bostroms argument doesn’t take into account exponentially fewer resources at each layer of simulation - You don’t just get to invoke infinite resources unless you can explain it - Pg 232 FOR - The key idea with Achilles and tortoise in zenos paradox is that abstractions may not resemble at times. Frequently they are a good match, but sometimes they are not, as we see in pg 249 - Question pg 250 (FOR) - [https://podcasts.apple.com/us/podcast/the-theory-of-anything/id1503194218?i=1000681411903](https://podcasts.apple.com/us/podcast/the-theory-of-anything/id1503194218?i=1000681411903) 1:02:44 - Review: can we render a logically impossible  environment in virtual reality? - Principle: never hold an explanation immune from criticism. That prevents error correction.  - You need to get better at getting _excited_ when there is something that requires updating your world view.  - Theory: if the AI is just “hype” then I definitely have ~6 months to sort out all my ideas about philosophy. They aren’t going anywhere. - Becoming a specialist - should I have stayed focused on what is science, popper, problem solving? No when you get to BOI you will get a great review of that ! - Use more first person in my essay? include myself as part of the story? - [Episode 83: Popper's Second Axis (aka Bruce's Epistemology?)](https://podcasts.apple.com/us/podcast/the-theory-of-anything/id1503194218?i=1000652484129&t=4688) - * Bruce argument 0 * One refutation of solipsism (or at least specific instances of it) is that there is always an experiment that you could design to test that you were in a simulation. The only way that this wouldn't be the case is if the simulation was perfectly internally consistent. Otherwise, you could eventually conjecture and test for the inconsistencies. The rendered environment would also have to be such that no explanations of anything inside would ever require one to postulate an outside. The environment, in other words, would have to be self-contained as regards explanations. So when someone argues for solipsism, you are arguing that the designers would have been able to construct a simulation with 100% accurate consistency.They are assuming perfection by the designer. Because we are fallible this would be very hard if not impossible. However, the [Autoverse](Autoverse.md) may be an example of otherwise. * So the real philosophical mistake that people are making is that we live a simulation so complete and internally consistent that there is no need to conjecture that there is an outside world * Bruce argument 1 * The *morass* is a giant infinite bundle of explanations that are indistinguishable from each other because they cannot be checked or tested. The morass is a great place to find interesting conjectures. But you have to take the conjecture out of the morass and make it *checkable* or *testable* before it can be distinguished between every other explanation inside the morass. * There is actually no reason to take the stance that we live in a perfectly consistent simulation run by gods, precisely because such a simulation is perfectly explained by realism, by definition. Even by the logic of solipsism itself. * There is no point in taking the stance that we live in a perfectly self consistent simulation, precisely because a perfectly self consistent simulation is actually explained by realism. Thus realism is a preferable theory because it constrains more and therefore explains more. * In fact it is hard to conceive how we could get the simulation hypothesis out of the morass, because it is intentionally framed to never come out of the morass. * Bruce argument 2 * Myth of the Framework, pg 60: By contrast, the correct method of critical discussion starts with the question "what are the consequences of our thesis or theory?" - "A kind of critical discussion does not seek to prove or justify or establish a theory, but tries to test a theory under discussion by finding out whether its logical consequences are all acceptable." * In other words, Deutsch attacks solipsism on both of poppers axes (TOA ep 83). Deutsch first shows that the defenders could have made their theory testable, but then it was have been refuted or at least criticized out of existence. But then he showed that their theory had no testable form in the vague-maned version. * This is the correct way to apply critical rationalism (and we can see that it applies to both scientific and philosophical theories) * Bruce argument 3 * the central aspect of poppers epistemology is: how do you chose to formulate your theories? You must do so to take greater risk * Mitchel ml book, epistemological claims  - Ram past the hard problem of consciousness  - [https://x.com/t_andy_keller/status/1899154774227878250](https://x.com/t_andy_keller/status/1899154774227878250) - Is down ward causality a thing (in godels theorem?) - Logical possibility vs physical possibility - diagonal argument is quite powerful! - Add deutschs view on mathematical calculations from Claude (vr is really about rendering some set of intrinsic rules-those could describe a physical system, an abstract one, and so on) - you just want the right initial seeds. Everything else should flow from there. Your mind will ask the right questions, you just need to know where to start  - How can we create theories with higher _content_ - From deutsch: a creative algorithm is really about the ability to create new problems. - A reductive theory of everything would describe and predict the copper atom on Churchill nose, but not explain it! This is almost by definition, for a prediction is not an explanation.  - Counter factual are closely tied to explanation  - The way you frame the question is the main problem! This was key in zge synthetic data (eventually could frame in terms of sensitivity to inputs). I think this is also discussed in Hazlitt book? Or somewhere else you read about framing questions and how you view the problem and what Lens’s you focus on? - Something ai doesn’t do is come up with clear through lines - it just scatters a ton of stuff. It’s not understanding. My writing has shown this - - https://www.incrementspodcast.com/24 - 37:00 - increments pod cast on poppers third world (dig into the idea of “consequences”) - 20:00 “what is” questions - we want to avoid that. But how doesthis relate to creativity. What is the role of what is questions? - [https://www.incrementspodcast.com/59](https://www.incrementspodcast.com/59) - Induction refutation - bar argument - ZGE - are we dependent on small probabilities? Could we avoid that? - Taleb survivorship bias (could this be at zge)?