Artificial intelligence (AI) is rapidly advancing, with tech giants like Google and OpenAI leading the charge. Despite significant investments in AI infrastructure, achieving Artificial General Intelligence (AGI) – where AI can think and reason like humans – remains a distant goal. Alexander Embiricos, the lead of OpenAI’s Codex, has pointed out that the development of AGI is hindered not by the AI models themselves but by the typing speed of humans.
Embiricos, speaking on Lenny’s Podcast, emphasized that the current bottleneck in AGI progress lies in human typing speed and manual prompt management. He highlighted the critical role of “human typing speed” and “human multi-tasking speed on writing prompts” as the underappreciated limiting factor in AGI development.
According to Embiricos, the reliance on humans to write prompts and validate AI-generated output slows down progress significantly. He stressed the need for AI agents to review work instead of humans to expedite advancements. Embiricos suggested restructuring AI systems to relieve humans from prompt writing and validation tasks, advocating for agents to be inherently useful to boost productivity rapidly.
While acknowledging the complexity of transitioning to fully automated workflows tailored to specific applications, Embiricos expressed confidence in imminent progress. He predicted that early adopters would experience substantial productivity gains, followed by widespread automation by major tech companies. This shift towards AI automation is seen as the gateway to unlocking AGI potential.
Embiricos anticipated a wave of productivity growth, likening it to a hockey stick graph, with early adopters leading the way and larger companies following suit. As productivity spikes through AI automation, Embiricos envisioned a path towards AGI realization. The envisioned surge in productivity is expected to fuel advancements in AI labs, paving the way for AGI breakthroughs in the near future.