Korva is a scroll of the internet’s best ideas.
Zuckerberg publishes a 6,500-word AI essay
The essay, subtitled The Path to a Positive AI Future, describes an optimistic AI worldview. Meta would release open-source models, create an agent for everybody, make it free, and give everyone in the world their own tutor.
It calls superintelligence invention rather than automation and holds that AI safety should rest on a balance of power, with no singular centralized intelligence. It asks why the people who call AI dystopian are rushing to create it, and it calls the notion that AI is so dangerous that the only safe path is an extreme concentration of power inherently problematic.
On jobs, the essay states that people fear automation will outpace individuals' capability growth, and that there is no rule that AI must increase automation faster than it increases individuals' capabilities or demand for new skills. It also defends the principle that you can learn from anything you can observe, stating that all AI models are derived from human knowledge.
How do open-weight models differ from open source?
An open-weight model publishes its trained parameters so anyone can download and run it. Traditional open-source software publishes the full source code under licenses that allow reuse and modification. Open-weight releases usually leave out the training data and training code, and some carry usage restrictions. Meta's Llama 2, released in July 2023, came with a license that allowed commercial use under conditions.
Why does one-on-one tutoring work so well?
Educational psychologist Benjamin Bloom reported in 1984 that students tutored one-on-one with mastery learning performed about two standard deviations better than students in conventional classes. That put the average tutored student above roughly 98 percent of the classroom group. Tutoring lets instruction adjust to each student's mistakes at once. Bloom called the challenge of matching that result at scale the 2 sigma problem.
How have past machines affected jobs?
Economist James Bessen found that the number of US bank tellers rose after ATMs spread, from the 1970s through 2010. ATMs cut the staff needed per branch, which lowered the cost of running one, so banks opened more branches. Tellers shifted toward sales and customer service. The case is a standard example of automation raising demand for the service it automates.
More from All-In with Chamath, Jason, Sacks & Friedberg
