Until a year ago, Artificial General Intelligence, or AGI, which allows computers to have human-like general intelligence, was thought to be the realm of fantasy, an idea which could possibly come to fruition in a few decades. Cut to 2023, and early Artificial General Intelligence might already be available around the world for $20.
A new paper has suggested that GPT-4 could be viewed as an early, but incomplete version of AGI. “Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system,” the paper says.
Artificial General Intelligence (AGI) is a term used to describe a type of artificial intelligence which is characterized by its ability to understand a wide range of tasks and processes, much like a human would. AGI is seen as a more general form of AI, and it is the idea of creating an AI machine that is capable of performing any intellectual task that a human can, with the same level of sophistication. Unlike narrow AI, AGI would be able to reason, plan, learn and problem solve. AGI could potentially have access to the same level of knowledge as a human and be able to apply it to any given situation.
The new paper, by Bubeck et al, now says that GPT-4 is approaching the AGI horizon. “In this paper, we report on our investigation of an early version of GPT-4, when it was still in active development by OpenAI. We contend that (this early version of) GPT4 is part of a new cohort of LLMs (along with ChatGPT and Google’s PaLM for example) 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. Moreover, in all of these tasks, GPT-4’s performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT. Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system,” the paper says.
The paper studies GPT-4’s performance on several tasks, including coding, mathematical abilities, interaction with the world, interaction with humans and discriminative abilities of GPT-4. The researchers found that it performed remarkably well. This expertise, across a variety of fields, is signs of early Artificial General Intelligence. While most bits of software perform specific tasks — Gmail sends email, and Amazon allows people to buy things online, ChatGPT can perform several different kinds of roles with high proficiency, indicating it’s coming close to what human intelligence is.
And it’s hard to overstate the implications of having created an AGI. An AGI could immediately make most white-collar human jobs redundant. More interestingly, if married to advances in robotics, it could even end up performing blue-collar jobs, at which point it might be indistinguishable from real humans. These were fantastical concepts just a year ago, but with GPT-4’s launch, we might be at the cusp of a brave new world.