Tags
All-In PodcastDavid FriedbergDavid SacksJason CalacanisChamath Palihapitiya
Date Consumed
October 12, 2022
Action
Count (Month)
Created
Oct 12, 2022 7:35 PM
Last Edited
Oct 12, 2022 8:46 PM
60 year cycle
- humans sense environment, generate knowledge, build model that predicts an outcome, and use it to drive actions
- same for computers, which are evolving how they build models
- initially, basic algorithms — completely deterministic, static models built by humans (input > output) — produces static solutions
- next, data science — as data generation proliferated (more/cheaper/better sensors, transmission, storage and compute — deterministic, static models could have parameters resolved by more data being generated — produces more predictive, better static solutions
- next, machine learning — enabled parameters within static models to become dynamic — model was static, but within the model the parameters became dynamic as more data came into the system and was updated — produces more predictive, dynamic solutions
- next, artificial intelligence — there is so much data, that you can resolve the model itself (not just parameters) from the data…i.e. the algorithm can be written by the software, resulting in dynamic models and dynamic parameters — so the software dynamically changes the model and the parameters — produces more predictive, better dynamic solutions
- i.e., statistical models are now resolving dynamically, and as a result they are far more predictive, and as a result we see far more human-like behavior in systems (e.g. better Chess/Go engines, self-driving systems, etc.)
- Eventually the algorithm becomes more complex than any human could have written, and suddenly the AI has built its own intelligence and is able to be predictive in a way that no human could have done
- All based on statistics…nothing new, other than techniques that put this into practice
- Geometric increase in data + geometric decline in cost to generate, transmit, store, and compute data
- If we solve enough specific AI problems, general AI will emerge
With AI, the marginal cost of intelligence will go to zero
- for humans: we must become more human (only way not to compete with robots)
Human transitions
- passives — consume work of the land, then
- create tools, become laborers — plow fields, cook, etc., then
- create machines to automate labor, become creators — knowledge work to create intelligence via computers, then
- create artificial intelligence, become narrators — knowledge work gets supplanted by AI (software), so we direct / dictate to the AI
- e.g. instead of creating the blueprint of a house, we describe it to AI and it creates the blueprint, orders materials, coordinates machines to build it, etc.
- e.g. instead of creating a movie, we dictate the movie we want to AI and iterate through the creative process
- e.g. instead of creating or buying a video game, you narrate the game you want to play and AI gives it to you
