AI Startup Founders Debate the Creation of Artificial General Intelligence

TL;DR

A group of AI startup founders debate the timeline and definition of AGI, with opinions ranging from "already here" (GPT-3 qualifies) to "very far away" (current models are just pattern matching, not intelligent). The core disagreement hinges on whether AGI requires consciousness/sentience or is just a capability spectrum, and whether GPT-4's apparent logic skills constitute genuine reasoning.

Key insights

• Definitional ambiguity is central: AGI as "high-level machine intelligence" (human-level or better on all tasks) vs. AGI as an autonomous agent with consciousness and beliefs.
• Several founders argue current large language models (e.g., GPT-4) are not reasoning but simply predicting the next token via statistical pattern matching.
• One founder asserts GPT-3 already qualifies as AGI; another says AGI is a slow-moving continuum we are already in.