Sam Altman Shifts Strategies in Pursuit of Artificial General Intelligence Achievement
In this rewritten piece, we delve into the intricacies of Sam Altman's latest take on Artificial General Intelligence (AGI). Often overlooked by the media, Altman's insights are invaluable due to his reputation as a trailblazer in the AI community. But before we dive into the heart of the matter, let's ponder the significance of cheese.
Cheese and Goals: The Mouse and the Cheese Maze
You might be familiar with the classic line, "Who moved the cheese?" This aphorism stemmed from a popular 1998 book by Spencer Johnson. The book encouraged introspection about goals and seeking their attainment, using the metaphor of a mouse trying to find cheese hidden in a maze. The cheese symbolizes a goal, so when the goal changes, it's essential to adapt.
Relating this to AGI, the definition of AGI will determine the direction we take to achieve it. Imagine if AGI were defined as blue with polka dots. Developers would strive to create AI systems with those characteristics, believing they've achieved AGI. But what if AGI should truly be defined as orange with big stripes? Pursuing the wrong goal can be fruitless.
It's common to assume that the AI community has a clear and definitive understanding of what AGI entails. Unfortunately, we must confront the truth: no universally accepted AGI definition exists. Various groups can propose their definitions, turning the AGI landscape into a chaotic, ever-shifting maze.
The Slippery Slope of AI, AGI, and ASI
I previously addressed the ambiguity surrounding terms such as AI, AGI, and Artificial Superintelligence (ASI)[1]. I'll summarize the critical points here for brevity:
- AI originally referred to any machine or computer system that exhibited human-like intelligence. However, as the field developed, the term became loosely applied to everything from simple software to cutting-edge servers.
- Expert systems or knowledge-based systems gained prominence in the 1980s, often narrowly focused on specific tasks.
- The term AGI was coined to represent the top echelon of AI, surpassing the limitations of expert systems. AGI would exhibit both narrow and general intelligence, afoot with human intelligence in all respects.
A Tighter, More Uplifting AGI Definition
Let me present three definitions of AGI for your examination. We'll start with the most ambitious:
- "AGI or artificial general intelligence is an AI system that exhibits intelligent behavior of both a narrow and general manner, on par with that of humans, in all respects."
This is a challenging goal, encompassing both narrow and general intelligence that mirrors human performance in all facets of cognition. Let's take a closer look at two looser definitions:
The OpenAI AGI Definition
OpenAI's mission statement pins down AGI as "highly autonomous systems" outperforming humans in economically valuable work[2]. It's essential to note that this AGI definition involves only the most economically valuable tasks.
An Openly Wider AGI Definition
Sam Altman, CEO of OpenAI, proposed the following more relaxed definition: "AGI is a weakly defined term...a system that can tackle increasingly complex problems, at the human level, in many fields."
Comparing these definitions, it becomes clear that the more stringent definition of AGI is more difficult to achieve. But it is worth aiming for the stars, even if we only reach the moon.
[1] Link to the previous AI definitions article
[2] Link to the OpenAI Charter statement webpage
- Sam Altman, with his reputation as a trailblazer in the AI community, provides a more relaxed definition of AGI as a system that can tackle increasingly complex problems at the human level in various fields.
- In his take on AGI, Altman emphasizes the importance of having a clear definition, avoiding situations where developers might mistakenly believe they've achieved AGI if the definition doesn't align with the actual capabilities of the AI system.
- The ambiguity in defining AGI is a challenge that the AI community faces, as there is no universally accepted definition, leading to varying interpretations and goals in the development of AGI.
- The lack of a clear-cut definition for AGI can have legal and regulatory implications, as stipulations and laws may struggle to keep up with the rapidly evolving field of AI, potentially leading to inconsistencies and uncertainties in its regulation.
