| … | … | @@ -12,4 +12,62 @@ Social media, via David Marx @digthatdata.bsky.social |
| 12 | 12 | * unpopular != wrong |
| 13 | 13 | * attention grabbing != important |
| 14 | 14 | |
| 15 | Bennnet: |
| 16 | * Valence (attraction, repulsion) |
| 17 | * "Weakness" vs "simplicity": Weak systems are those that allow more |
| 18 | variety of output than strong systems. Weak systems outperform simple |
| 19 | systems (by a lot) in training/fitting data. |
| 20 | * Weak tools can be employed in a greater variety of tasks. |
| 21 | |
| 22 | Paper: |
| 23 | Investigating the impact of trial retractions on the healthcare evidence ecosystem (VITALITY Study I): retrospective cohort study |
| 24 | BMJ 2025; 389 doi: https://doi.org/10.1136/bmj-2024-082068 (Published 23 April 2025) |
| 25 | Abstract |
| 26 | Objective To investigate the impact of retracted trials on the |
| 27 | production and use of healthcare evidence in the evidence ecosystem. ... |
| 28 | |
| 29 | So this is a specific deep dive into how bad info pollutes the |
| 30 | evidence-based information ecosystem. |
| 31 | |
| 32 | Manuscript Notes |
| 33 | ---------------- |
| 34 | Notes taken while reading: |
| 35 | Are Biological Systems More Intelligent Than Artificial Intelligence? |
| 36 | Michael Timothy Bennett |
| 37 | https://arxiv.org/abs/2405.02325 |
| 38 | |
| 39 | Section 3.1 of version 4 of the PDF appears to contain a key typo that |
| 40 | makes it difficult to understand. It says that a "vocabulary" is a |
| 41 | subset of the set of states. Version 3 correctly defines the |
| 42 | vocabulary as the a subset of the power set of states. |
| 43 | |
| 44 | Recapping the definitions: |
| 45 | * There is a set of "states" Phi. |
| 46 | * A "program" is a subset of Phi. |
| 47 | * A "vocabulary" is a set of programs. It is fixed externally, a |
| 48 | priori, by the reader/external observer. |
| 49 | * A "language" is the set of all(?) subsets of the vocabulary, such that |
| 50 | each subset has a non-empty intersection. That is, each element of |
| 51 | the language is a set of programs, such that there is at least one |
| 52 | state is common to all programs. |
| 53 | * The definition is not clear; he may be defining the "language" to |
| 54 | be the union of all ultrafilters. Or he may be defining a language |
| 55 | to be the union of just some filters. |
| 56 | * A statement is an element of the language. The extension of a |
| 57 | statement is the filter containing that statement. |
| 58 | * The "weakness" of a statement is the cardinality of it's extension. |
| 59 | * ... definitions are a bit garbled and riddled with typos that make |
| 60 | comprehension difficult. |
| 61 | |
| 62 | I dunno. This paper tries to formulate hierarchical systems in terms of |
| 63 | filters (as in topology) and boolean algebras thereon, using |
| 64 | non-standard terminology, but I just don't see where it goes. If I blur |
| 65 | my eyes, this all just collapses to pretty standard topology. The |
| 66 | problem is that general topology has many far deeper and far more |
| 67 | sophisticated statements of which above is a scraping of the surface. |
| 68 | Alas. |
| 69 | |
| 70 | |
| 71 | Other stuff |
| 72 | ----------- |
| 15 | 73 | TBD. I am going to write the next blob in Lyx. See where that goes. |