OpenAI Unveils Scale to Track AI Progress: ChatGPT Starts at Level 1
The rapid rise of AI has made it a defining feature of modern technology. As these systems become more complex, there’s a growing need to gauge their capabilities. To address this, OpenAI developed an internal scale to track how close their large language models are to achieving human-level intelligence (AGI). This framework essentially measures how “smart” their AI tools are.
OpenAI created a ranking system with five levels to measure how “smart” their AI tools are, according to a Bloomberg report. This framework compares AI capabilities to human problem-solving skills. Currently, their well-known ChatGPT falls under Level 1. Interestingly, OpenAI believes ChatGPT is on the verge of reaching Level 2, which signifies problem-solving abilities on par with someone holding a PhD. As we move up the scale, Level 3 AI can act independently for users, while Level 4 showcases creativity by generating new ideas.
The ultimate goal of this ranking system is Level 5, representing Artificial General Intelligence (AGI). AGI, as defined by OpenAI, is a super-smart system capable of tackling most valuable tasks better than humans. This framework hints at OpenAI’s ambition to develop tools that reach this pinnacle of intelligence. It aligns with CEO Sam Altman’s recent push for AGI development, regardless of the cost.
Interestingly, Bloomberg also reported that OpenAI’s latest model, GPT-4, showcases reasoning abilities similar to humans. This internal ranking system follows news of their upcoming LLM, GPT-5, which the CEO promises to be a breakthrough.
There’s a buzz around OpenAI’s upcoming language model, GPT-5. While details remain scarce, CEO Sam Altman recently expressed optimism about its potential at the Aspen Ideas Festival. He acknowledged that the model is still under development but aims to fix some of the shortcomings found in GPT-4.
Also read : Musk says Neuralink aims to cure paralysis and memory loss.
This excitement follows the impressive launch of GPT-4o in May. GPT-4o surprised us with its mathematical skills and ability to hold emotionally intelligent conversations, setting a high bar for its successor, GPT-5.
