Are yet-unknown ideas and yet-undiscovered facts in between the known or outside of the known?
What should those even mean, conceptually?
Should we look closer, or should we look farther?
I need to get as fast at prototyping and validating deep learning models as I am at prototyping and validating web apps. This requires investment in:
- Knowing/understanding a few tools deeply and intimately
- Understanding core concepts deeply
- Investing in custom tooling and infrastructure where it makes sense
Generative AI technologies like GPT and DALL-E could allow a "creator economy for worldbuilding" to exist.
getting so good at life you get by without a job, you're just in demand and people want you around unblocking them, performing miracles and shit
The foundations of modern physics were laid rapidly, by a small group of brilliant people collaborating loosely and autonomously, in a few decades in the early 20th century. It bothers me that we haven't seen our understanding of nature advance so quickly after WWII. So I've been thinking about what the necessary conditions may be to create such a rapid, creative environment for discovery and progress. Here is my current hypothesis:
- A small group of the best people
- working in close collaboration and communicating frequently, but pursuing their hypotheses independently
- on a few, well-known important problems onto which their entire field is focused.
I hypothesize that similar conditions were true for computing in the late 60's/early 70's and similar conditions may be true now for ML. Loose collaboration between independent leaders across a focused field.
How might we create and nurture such conditions?
Lex Fridman's recent interview with Brendan Eich reminded me of Flash, which seemed relevant to my recent writing about inventing new software materials and laws of physics when designing interfaces. I've heard Flash, especially the interactive animation editor, praised so often for its interface innovations/interface metaphors. I should look into it more.
I think I would pay up to $20/mo for a service that would take a Twitter handle or email address of a stranger, and based on some cursory Google searches and Internet sleuthing, gave me a few-paragraphs summary of who that person is, and why they might (or might not) be interesting. I have some friends who are very good at this, but it strikes me as the kind of thing that may not be too difficult to automate and scale with language models.
Our brains (visual systems, language centers, etc.) are capable of perceiving, internalizing, and thinking in terms of a vastly larger set of primitives than we currently use them for, and we probably give them credit for.