OpenAI’s push for human-level AI is underway

Reports of OpenAI’s move towards human-level artificial intelligence (AI) have the business world abuzz with visions of AI-powered commerce that could rewrite the rules of global trade – if the technology lives up to the hype.

General artificial intelligence (AGI) represents a hypothetical leap beyond current AI, aiming to match or exceed human cognitive abilities in various tasks. In commerce, experts say AGI could revolutionize decision-making, providing deep insights into market trends and consumer behavior. It promises to streamline operations, from supply chain management to product development, with unprecedented efficiency.

According to Reuters, OpenAI is developing a new approach to its AI models under the codenamed “Strawberry” project, which aims to improve the advanced reasoning capabilities of the models. While the exact details and timeline are unclear, the initiative focuses on improving AI’s ability to understand and process complex information, positioning the Microsoft-backed startup to push the boundaries of what its models can achieve.

“OpenAI’s pursuit of human-level reasoning is not just a technological marvel; it is a narrative of pushing boundaries and igniting new possibilities in every sector.” Ghazenfer Mansoor, founder and CEO of Technology Rivers, told PYMNTS. “In business, AI can dramatically change how supply chains are managed, predict market trends with great accuracy, and make customer experiences highly personal on a massive scale.”

OpenAI did not immediately respond to a request for comment from PYMNTS.

Human-like reasoning

Earlier this year, Reuters reported that OpenAI staff showed demonstrations of AI models capable of answering complex science and math questions, with one model scoring over 90% on a championship math dataset. At a recent internal meeting, OpenAI showed off a project with new human-like reasoning capabilities, but it’s unclear if this is related to the Strawberry project.

“The way such an algorithm might work is by generating multiple options, following a tree of possibilities, and then reasoning about the outcome and choosing the best path.” SmythOS CTO Alexander De Ridder said of PYMNTS. “This is similar to how chess players think several steps ahead before choosing to move their piece.”

De Ridder suggested that OpenAI’s innovation is likely to involve “an algorithmic discovery of how to do this efficiently and at scale,” potentially combining “autonomous web research and the use of tools to arrive at a reasoning discovery.”

OpenAI has RECITED developed a five-level classification system to track progress toward AGI, placing its current technology at Level 1. The company, known for its advanced language models, plans to share this framework with investors and outside parties, signaling a potential breakthrough in AI research that can outperform humans in various tasks.

Contextual reasoning is critical to the advancement of AI technology, so being on the verge of bringing this capability to life brings us another step closer to general artificial intelligence. Nick Sanchezsenior solutions architect at DRUID HEsaid of PYMNTS.

He added, “By including reasoning as a dimension within predictions, AI can contextualize the intent of the question being asked and frame the data in a way that best suits whoever is asking it.”

From finance to health

Experts highlight various applications of advanced AI reasoning. “Creating robotic assistants that can move in the real world and reason in it would be a significant advance. Reasoning about scientific data at the human level would transform and accelerate the way we make discoveries,” noted De Ridder.

Sanchez focuses on the human-centric implications: “Advanced reasoning will be able to create more context of the initial input, which influences the outcome, creating a more efficient and personalized interaction for the user.” He gives an example: “If someone asks their banking chatbot about their account balance because they have a bill to pay tomorrow but no funds, the AI ​​can share a more empathetic response because of a better understanding of the situation. “

The potential impact on health care is particularly significant. Mansoor noted, “AI with human-level reasoning can revolutionize diagnosis, treatment planning and drug discovery. It can analyze complex medical data, understand patient stories and propose personalized treatment plans with unprecedented accuracy.”

“AI-powered thinking can increase efficiency, help with data-driven decisions and encourage new ideas. Companies using AI can change the way they do things, becoming faster to adapt and better at planning for the future,” added Mansoor.

On the competitive landscape, De Ridder said: “OpenAI has lost its lead in being the best in every category and for every type of user, but it maintains perhaps the strongest model in the world and leads in accessibility.” He pointed to advances by other companies, including Anthropic’s Claude 3.5 family of models and Google’s Gemini 1.5 pro flash.

The race for human-level AI also raises important questions about comparison and evaluation. De Ridder explained: “It is universally accepted that current standards need improvement. Previous benchmarks have either been tainted by new training, proved inadequate/flawed, or become too easy for AI models.”

The potential for transformative applications increases as the AI ​​industry pushes toward human-level reasoning. However, challenges remain in bridging the gap between current AI models and true human-like understanding and reasoning. Ethical considerations, including privacy concerns and the potential for job displacement, also loom large.

“As these capabilities improve, they can change many fields, give businesses new and profound insights and usher in a new era of creativity and effectiveness,” Mansoor said, stressing the need for responsible development and deployment of these powerful technologies.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.