OpenAI’s Codex lead Alexander Embiricos says human interaction limits are becoming a key bottleneck for AGI development.
He argues AI systems must be redesigned to work reliably by default to reduce constant human validation.
Early adopters could see sharp productivity gains as soon as next year, with broader automation paving the way for AGI.
Big tech companies like Meta, Google, and OpenAI are racing to develop artificial general intelligence (AGI). While the targets for achieving that appear to be far higher than expected, senior executives at these companies have been quite vocal about it as the industry moves towards AI. A senior OpenAI executive recently stated that human limitations such as typing speed and the time required to review AI-generated work could be one of the most significant impediments to the arrival of AGI.
SurveySpeaking on Lenny’s Podcast as cited by Business Insider, Alexander Embiricos, who heads product development for OpenAI’s coding agent Codex, said progress toward AGI is increasingly constrained not by models or computing power, but by how quickly humans can interact with and supervise AI systems. As AI agents become capable of handling larger volumes of complex work, the need for humans to constantly write prompts and manually validate outputs is emerging as a key bottleneck, he argued.
Embiricos stated that while AI agents can already observe and assist with tasks such as coding, productivity gains are limited if humans must manually review every output. According to him, if we want to unlock the next stage of growth, systems must be redesigned so that agents become reliable by default, reducing the need for constant human intervention.
He stated that this shift is critical for triggering hockey stick productivity growth, which happens when gains are gradual for a period of time before accelerating sharply. According to Embiricos, early adopters may see a boost in productivity as soon as next year, with larger organisations following suit as workflows become more automated.
While he cautioned that there is no single solution for fully autonomous AI systems and that different use cases will necessitate tailored approaches, Embiricos predicted that AGI will emerge during the transition period between early productivity gains and widespread enterprise automation.
Follow Us
Ashish Singh
Ashish Singh is the Chief Copy Editor at Digit. He's been wrangling tech jargon since 2020 (Times Internet, Jagran English '22). When not policing commas, he's likely fueling his gadget habit with coffee, strategising his next virtual race, or plotting a road trip to test the latest in-car tech. He speaks fluent Geek. View Full Profile
