AI is having a moment. Decades of fundamental discoveries and years of quiet revolutions are giving rise to a sudden explosion of applications in unexpected places.
Unlike other hype-heavy technologies of recent years, AI is already useful. While it’s acceleration might taper off, the genie isn’t going back into the bottle. This leaves us with an infinitude of questions to grapple with, by people of all levels of technical literacy, which will impact generations to come. That is why this newsletter exists: to inform the curious and help more people understand the underlying principles and possible outcomes of our collective transition towards artificial general intelligence.
You don’t have time to learn everything but want to keep up to speed. Every week you will find a selection of links with interviews and explainers by players in the AI space, with personal commentary from somebody dedicated to exploring the intersection of human and machine intelligence. The AI revolution might very well eclipse the technological advancements of the last century, or it could go nowhere. This will help you navigate whatever happens next.
MZ
Sebastian Bubeck speaks about Sparks of AGI 🏳️ (50 min)
This lecture by a Microsoft AI researcher sparked my wanting to do this newsletter. Early experiments with GPT-4 show a limited ability in generalizing its problem-solving approach. In other words, GPT-4 is starting to demonstrate the capacity of broadening its abilities and act out unexpected types of intelligence and behaviors. Highly recommended.
Alondra Nelson and Ezra Klein 🏛️ (1h)
US Policy advisor Alondra Nelson talks to Ezra Klein about how the public could benefit from an Artificial Intelligence Bill of Rights.
This is a sui generis technology. We haven’t really dealt with anything like it. The harms are unforeseeable. The technologies, the systems are uninterpretable.
Greg Brockman on TED 2023 🔧 (30 min)
OpenAI cofounder explores the underlying design principles of ChatGPT, demonstrates the power of plugins, extrapolates about the nature of superintelligence and answers questions about emergence by Chris Anderson.
35 ways real people are using AI right now 🥌 (5 min)
Artificial intelligence models are finding their way into surprising places.
Sam Altman on Lex Fridman 🔓 (2h 20 min)
OpenAI CEO Sam Altman discusses the potential dangers & ethical considerations of AGI development, emphasizing the importance of safety, user control, and shared responsibility in shaping the future of AI technology. Bonus points for Sam not having watched (or wanting to talk about) Ex Machina (!).
John Carmack's different path to AGI 👀 (20 min)
Rare and insightful interview with John Carmack about why he is taking on AGI as his next big technological bet. Highly recommended.
I see the destination. I know it’s there, but no, it’s murky and cloudy in between here and there. Nobody knows how to get there. But I’m looking at that path saying I don’t know what’s in there, but I think I can get through there—or at least I think somebody will. And I think it’s very likely that this is going to happen in the 2030s.
Paper: Sparks of AGI ⚡️ (155 pages)
Companion paper to the lecture above by Sébastien Bubeck which I am still trying to wrangle.
We contend that (this early version of) GPT-4 is part of a new cohort of LLMs […] that exhibit more general intelligence than previous AI models. We discuss the rising capabilities and implications of these models. We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting.
Emerging Vocabulary
Alignment
The process of ensuring that an artificial general intelligence (AGI) system's goals, values, and behaviors are in harmony with those of its human creators and users. It is crucial for achieving AGI because without proper alignment, AGI systems might pursue objectives that are not beneficial or even harmful to humanity.
Project Showcase
Metaphor
Link prediction search engine designed from scratch using AI. Works by extrapolating on a prompt with probably links and resembles how someone on the internet might refer to links. [Via Mia]
If Artificial Insights resonates with you, please help me out by:
You are receiving this newsletter because you signed up on envisioning.io.
Artificial Insights is a weekly newsletter about what's happening in AI and what you should understand about our transition toward AGI. Each issue features interviews, articles and papers selected for an audience of leaders and enthusiasts.