Generative AI's greatest advantage is visualization. It closes the implementation gap by turning abstractions into explorable explanations.
Generation is the default benchmark because it is easy to measure... And is "productive" (if that has meaning) but is inefficient.
Understanding is much harder. The challenge today is to build systems that make ideas more coherent, inspectable, and transferable.
Using Star Trek as a template, ask if will we build more Vulcan-like technology, using intelligence to expand knowledge and insight, or scale Borg-like accumulation and assimilation.
Submissions
Tesler's Law
en.wikipedia.org
11 points·by soupspaces··3 comments
Fabric Interfaces Tutorial [video]
youtube.com
2 points·by soupspaces··0 comments
Deep Brain Stimulation
en.wikipedia.org
3 points·by soupspaces··0 comments
Eewoc: Extended-Reach Enhanced Wheeled Orb for Climbing
youtube.com
3 points·by soupspaces··0 comments
OpenAI Hacked Hugging Face
youtube.com
2 points·by soupspaces··0 comments
The future of AI – Eric Schmidt (2024)
youtube.com
1 points·by soupspaces··0 comments
Paper Soccer
paper.soccer
4 points·by soupspaces··2 comments
Eye Tracking on Your Cellphone
gazecapture.csail.mit.edu
10 points·by soupspaces··4 comments
Freeze-Casting
en.wikipedia.org
28 points·by soupspaces··6 comments
Docling - Composite file format
github.com
1 points·by soupspaces··0 comments
Mass Surveillance Wasn't an Accident
youtube.com
5 points·by soupspaces··0 comments
Test-time training 3D reconstruction
github.com
29 points·by soupspaces··2 comments
Cell Just Changed Biology
youtube.com
1 points·by soupspaces··0 comments
Information Awareness Office
en.wikipedia.org
1 points·by soupspaces··0 comments
Intro to Multiplicative Calculus [video]
youtube.com
1 points·by soupspaces··0 comments
Marble - (Education) Skill Taxonomy
github.com
1 points·by soupspaces··0 comments
Oneko
github.com
2 points·by soupspaces··0 comments
EEG Signal Quality of an In-Ear Wearable
pmc.ncbi.nlm.nih.gov
1 points·by soupspaces··0 comments
Perchlorate
en.wikipedia.org
5 points·by soupspaces··1 comments
Lean4Physics: Reasoning Framework for College Physics in Lean4
Except you can't possibly know that. New insight can arise regardless of whether mathematicians are trying to prove or disprove a statement, and regardless of whether the statement ultimately turns out to be true or false.
What to do about it? A reinassance of knowledge curators to swim against the current? A revolution slop salmon men? I doubt the incentives are aligned.
Generation is the default benchmark because it is easy to measure... And is "productive" (if that has meaning) but is inefficient.
Understanding is much harder. The challenge today is to build systems that make ideas more coherent, inspectable, and transferable.
Using Star Trek as a template, ask if will we build more Vulcan-like technology, using intelligence to expand knowledge and insight, or scale Borg-like accumulation and assimilation.