OpenAI's Grading System For AI Smartness: A Progress To 'Human-Level’ Problem Solving Skills

From Level 1 to Level 5, OpenAI can now judge itself.

Since the company launched ChatGPT, many things about AI that were etched to people's minds, are suddenly erased. The generative AI tool literally sparked an arms race, where other tech companies, large and small, are getting themselves in war for supremacy.

But generative AI products are just like any other AI that came before it, in which their abilities are somehow restricted to the data they have been trained on.

What's more, generative AI tools, despite becoming increasingly smarter, they still hallucinate from time to time.

OpenAI knows this very well, and this is why it created a tool to track the progress its Large Language Models are making toward Artificial General Intelligence

Revealed by Shaun Ralston on the social media platform X, the OpenAI API Dev Support said that at this time, OpenAI's ChatGPT is at Level 1.

Read: Paving The Roads To Artificial Intelligence: It's Either Us, Or Them

And so does all other modern chatbots for that matter. They are all still stuck at Level 1.

But here, OpenAI claims that it's nearing Level 2, which is a level where an AI system can solve basic problems at the level of a human with a doctorate-level education, or a PhD, but doesn’t have access to any tools.

OpenAI executives referred to AIs in this Level 2 as the "Reasoners."

The idea of having this kind of scale, is to see how AI progresses, and in order to know how far to set their skipping stones.

In this case, Level 3 refers to AI agents capable of taking actions on a user’s behalf, and Level 4 involves AI that can create new innovations.

Level 5, which OpenAI considers the final step to achieving AGI, is when an AI that can perform the work of entire organizations of people.

The five-level grading system is meant to see how far AI is underperformed, and soon-to-outperform humans.

The move marks OpenAI's effort to help people better understand its thinking about safety and the future of AI.

Read: AGI Is When AI Requires Less Training Data, But More On 'Reasoning Capabilities'

OpenAI has previously defined AGI as "a highly autonomous system surpassing humans in most economically valuable tasks."

Since the very beginning, the company has aims towards achieving AGI, and that Sam Altman for more than many times, have defined the importance of achieving it.

While the AGI is still a term that is quite far away, and that experts calculated that it may need billions of dollars worth of computing power to reach it, the timelines from the many experts vary wildly.

Some said that it's nearly impossible to reach AGI, and some said that it's possible, and it's just around the corner.

Back in October 2023, OpenAI CEO Sam Altman said humanity is "five years, give or take," before reaching AGI.

And to pursue that goal, the executive once said that he doesn't care if he burns $50 billion a year.