top of page
Gen-AI Employee Support & Automation Platform

OpenAI Nears Breakthrough in AI Reasoning Capabilities


According to Bloomberg and Reuters, OpenAI researchers believe they are on the brink of creating AI that can perform human-level "reasoning."


AI experts are divided on whether today's large language models (LLMs), which generate text and images, will ever fully understand the world and adapt to new information flexibly.


OpenAI has internally shared a framework defining five levels of artificial general intelligence (AGI):

1. Chatbots: AI with conversational language capabilities.

2. Reasoners: AI capable of human-level problem-solving.

3. Agents: AI systems that can take action.

4. Innovators: AI that can aid in invention.

5. Organizations: AI that can perform the work of an organization.


At a recent company meeting, OpenAI leaders stated their systems are currently at level 1 but are on the cusp of reaching level 2, "Reasoners."


Reuters detailed a project named "Strawberry," which aims to develop AI that can reason at a human level, plan ahead, solve multi-step problems, and independently conduct deep research. This project continues a previous effort known as "Q*."


Last November, rumours of a breakthrough at OpenAI circulated just before an unsuccessful attempt by the board to remove CEO Sam Altman. Altman hinted at significant progress during an Asia-Pacific Economic Cooperation event.


Despite these advancements, OpenAI's latest product, GPT-4o, combines text and visual modes but does not yet achieve human-level reasoning.


OpenAI's mission is to ensure AGI benefits all of humanity, but the concept still needs to be clarified. Some AI optimists believe AGI will evolve from today's LLMs as they expand and improve. However, skeptics argue that LLMs may not achieve human-level reasoning due to increasing resource demands.


OpenAI's five-level AI ladder mirrors frameworks used to understand progress toward fully autonomous vehicles. Google's DeepMind has a similar five-level breakdown for measuring AI advances.


Interestingly, neither OpenAI's nor Google's frameworks address the potential for AI systems to become autonomous or sentient.

bottom of page