Alexander Embiricos, Product Manager of OpenAI's Codex coding agent, has recently stated that a key bottleneck in developing advanced artificial intelligence is the speed at which humans can type.

In a podcast interview, OpenAI senior executive Embiricos argued that progress toward artificial general intelligence (AGI) is limited by the human role in the process. He explained that current systems require people to write detailed instructions, or prompts, and to manually review the AI's work, which creates a slowdown.

Embiricos also said the "current underappreciated limiting factor" is "human typing speed" or the speed at which people can multitask to create prompts, as per a report by Business Insider on the matter. He suggested the next step is to build systems where AI agents can autonomously validate their own work, removing the need for constant human oversight.

His view is that when AI can operate and self-correct without waiting for human input, it will lead to explosive "hockey stick" growth in productivity. Embiricos predicted that early adopters will begin seeing this surge in efficiency starting next year, with larger companies following in subsequent years, adds the report.

He proposed that the feedback loop from these productivity gains accelerating AI development itself could ultimately lead to the emergence of AGI, a hypothetical form of AI that matches or exceeds human reasoning.