The A.I. Dilemma - March 9, 2023
Center for Humane Technology
Imagine a world where artificial intelligence (AI) technologies have the power to reshape society as we know it—but at what cost? In this riveting video, Tristan Harris and Aza Raskin delve into the high-stakes race among AI companies to unleash their innovations, often without considering the safety nets needed to prevent catastrophic consequences.
But here's the catch: the AI landscape is evolving at breakneck speed, and this video is a testament to that. Filmed just four weeks ago, parts of their discussion were rendered obsolete within days by the release of the groundbreaking GPT-4. Don't miss this eye-opening conversation that captures a critical moment in the AI revolution, and join us as we explore the implications of transitioning to a post-AI world. Watch now and witness the future unfolding before your eyes!
Key Points and Takeaways
Talk Begins
The conversation opens by exploring how we tend to forget what we don't fully understand, often reverting to our comfort zone.
1:30 The Rubber Band Effect.
3 Rules of Technology
5:36 1. New tech, A New Class of Responsibilities
6:42 2. If a Tech confers power, it starts a race
6:47 3. If you don't coordinate, the race ends in tragedy
Exponential Improvement
Understanding exponential improvement is crucial. People typically struggle to grasp the concept of exponential progress, yet a confluence of factors guarantees that exponential improvement is inevitable. A.I. is advancing faster than we can comprehend.
14:13 Understanding the exponential improvement of A.I. and Machine Learning
Gollem-class AIs and Emergent Capabilities
The essential point is that A.I.s aren't particularly useful for their knowledge. They aren't well-suited for search tasks or generating bios because they tend to hallucinate and confabulate. It's more effective to provide them with facts and ask them to reason based on that information, though even then, verification is necessary.
Nonetheless, the most intricate models generate mental models of the world, exhibiting complex emergent behavior. This feature makes A.I. intriguing, useful, and potentially dangerous.
17:09 GLLMM: Generative Large Language Multi-Modal Model (Gollem AIs)
18:12 Multiple Examples: Models demonstrating complex understanding of the world 27:34 Total decoding and synthesizing of reality: 2024 will be the last human election [the politicians with the most computing power wins.]
29:55 Sudden breakthroughs in multiple fields and theory of mind
34:50 Gollem-class AI can make themselves stronger - A.I. can feed itself
38:40 Exponentials are challenging to understand
Feedback Welcome
The video contains a wealth of additional information, and I encourage you to watch it. While I don't necessarily agree with all the viewpoints presented - particularly those related to problem-solving - I believe this video effectively showcases the capabilities of new A.I.s and the speed at which they improve.
I created this post to spark conversations about A.I.s, their capabilities, and potential dangers. I look forward to connecting with you in the comments and via chat. - Nick
- Nick