Unsupervised Learning• Apr 9, 2026• 58:45Interview

OpenAI’s Chief Scientist on Continual Learning Hype, RL Beyond Code, & Future Alignment Directions

From Unsupervised Learning

Akko Paoki•Chief Scientist, OpenAI

Executive Summary

OpenAI's Chief Scientist, Akko Paoki, confirms the company is shifting its research focus from abstract benchmarks like mathematics to practical, real-world applications, believing current models are capable enough to materially impact the economy.
The company remains on track with its goals to develop a "research level intern" AI by September 2024 and a "fully automated AI researcher" by 2028, driven by advances in coding, mathematical reasoning, and a focus on creating autonomous, long-running agents.
OpenAI's strategy is heavily guided by "the bitter lesson," prioritizing scaling compute and general methods, with continual learning identified as the central goal of their current research path.
Paoki highlights the significant societal challenges ahead, including the automation of intellectual work, wealth concentration, and the governance of powerful AI systems, urging for a broader societal discourse on these issues.

10 quotes

Concerns Raised

The core AI alignment challenge of generalization remains an unsolved, long-term problem.
Large-scale automation of intellectual work will create significant societal challenges regarding jobs and wealth concentration.
The governance of extremely powerful AI systems, potentially controlled by very few people, is a new and critical question for society.

Opportunities Identified

AI models are now capable enough to materially change the economy and accelerate scientific discovery.
Developing autonomous AI agents that can work for days to complete complex research and engineering tasks.
Using AI's rapidly improving coding and reasoning capabilities to automate and improve AI research itself, creating a virtuous cycle of progress.

Key Themes

From Benchmarks to Economic Impact

OpenAI is strategically pivoting from using abstract, verifiable benchmarks like International Mathematical Olympiad (IMO) problems to focusing on the direct application of AI in economically valuable tasks and scientific research. Paoki states that models have reached a threshold of capability where they can now materially change how work is done.

This marks a significant inflection point in the AI industry, signaling that leading labs are moving from capability demonstration to actively deploying AI for real-world productivity and economic disruption.

The Path to Autonomous AI Agents

The development of long-running, autonomous AI agents is a primary focus. Paoki envisions an evolution from current coding assistants to systems that can work independently for days on complex tasks, such as AI research, to produce high-quality artifacts.

The successful development of such agents would represent a paradigm shift in knowledge work, automating complex intellectual labor and dramatically accelerating the pace of innovation.

AI as a Scientific Discovery Engine

The conversation highlights AI's emerging ability to contribute novel ideas to scientific and mathematical research, moving beyond mere pattern matching. Examples like the "first proof" challenge, where a model solved unpublished, research-level problems, demonstrate AI's potential for genuine discovery.

This capability could revolutionize science by providing researchers with powerful tools to generate hypotheses, solve intractable problems, and accelerate the discovery of new knowledge across various fields.

The Primacy of Scaling and Generalization

OpenAI's research philosophy is rooted in "the bitter lesson," which favors scaling compute and data over intricate, human-designed algorithms. This approach is central to achieving continual learning and solving the core alignment challenge, which Paoki frames as a problem of generalization in novel situations.

This philosophy dictates the allocation of immense resources towards compute and data, influencing the entire AI hardware and software ecosystem and defining the dominant paradigm for achieving more capable and aligned AI.

Societal and Governance Challenges

Paoki expresses urgency around the societal implications of advanced AI, particularly the large-scale automation of intellectual work, wealth concentration, and the governance of powerful AI controlled by small groups. He stresses that these are not just technical problems but require broad involvement from policymakers and society.

As AI capabilities advance rapidly, these societal and ethical questions become increasingly urgent, requiring proactive policy and public discourse to ensure a beneficial transition and mitigate potential negative consequences.

Get started free

Topics

AI Research OpenAI Akko Paoki Model Progress Scaling Laws The Bitter Lesson Continual Learning Autonomous Agents AI for Science Mathematical Reasoning IMO Problems AI Alignment Generalization Reinforcement Learning (RL)AI Coding Assistants Economic Impact of AI AI Governance

Processed Apr 15, 2026 yt-dlp + mlx-whisper + Gemini