San Francisco – OpenAI, the company that brought us ChatGPT, is embarking on its most ambitious project yet: building AI systems capable of conducting research entirely on their own. This new “grand challenge,” as the company calls it, aims to create artificial intelligence that can independently tackle complex problems currently beyond human capacity.
The ‘AI Research Intern’ and Beyond
The immediate goal, slated for September, is to develop an “autonomous AI research intern.” This AI would be tasked with a select number of specific research problems, acting as a proof-of-concept. However, the ultimate vision, projected for 2028, is a fully automated, multi-agent research system. Imagine an AI capable of generating novel mathematical proofs, uncovering new biological insights, or even devising solutions to intricate business and policy dilemmas – all without human intervention beyond the initial problem definition, which could be as simple as text, code, or even whiteboard scribbles.
A New ‘North Star’ Amidst Fierce Competition
This ambitious undertaking is not just about pushing technological boundaries; it’s a strategic pivot for OpenAI, now facing stiff competition from giants like Google DeepMind and Anthropic. With its previous breakthroughs in large language models shaping the AI landscape for millions, OpenAI’s next move is critically important. Chief Scientist Jakub Pachocki, a key figure behind GPT-4 and sophisticated reasoning models, sees this as the company’s guiding principle for the coming years. He envisions a future where “you kind of have a whole research lab in a data center,” with AI agents working tirelessly on humanity’s most pressing issues, from curing diseases to unraveling the mysteries of physics. This focus on long-term, autonomous problem-solving could redefine the pace of scientific discovery itself.
📰 Source: MIT Tech Review
