Former OpenAI Researcher Issues Grim Warning, Says Only 5,000 People Have Access to Frontier AI While 2 Billion Use It as ‘Better Google Search’
Former OpenAI researcher Andrej Karpathy warned that the gap between public understanding of artificial intelligence and its most advanced capabilities is widening, with a small group of professionals witnessing a rapid leap in what AI agents can accomplish.Billions Have Limited Exposure to Advanced AIOn Sunday, in a post on X, Karpathy said the gap in understanding AI capabilities is "widening," with billions of people barely exposed to large language models (LLMs) while others use advanced systems to complete complex tasks.He estimated that around six billion people have barely interacted with LLMs, while another one to two billion use free-tier AI tools casually, often treating them as "a better Google search" or a writing aid.By contrast, Karpathy estimated that roughly 20 million professionals, particularly in mathematics and software engineering, use frontier-grade models to complete projects that once took weeks or months.AI Agents Take on Massive ProjectsKarpathy said an even smaller group of about 5,000 people has internal access to the most advanced systems, which can coordinate thousands of AI agents on massive software projects, cybersecurity operations and mathematical research."The external world has seen the preview," he wrote, describing the capabilities of these advanced systems.He also warned that "human review and comprehension are starting to fall behind" as AI systems generate increasingly complex work.Karpathy attributed the widening gap to differences in access, task complexity and AI’s uneven performance across fields.He said AI capabilities are strongest in digital domains such as coding and mathematics, where results can often be verified.Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.– Napkin math somewhere around 6B people (~75% of the population) have barely come… Read Also: AI Agents Could Drive Up Costs and Computing DemandLast week, Space Exploration Technologies Corp. CEO Elon Musk backed Box Inc. CEO Aaron Levie’s warning that autonomous .Morgan Stanley estimated they could add $32.5 billion to $60 billion to the data-center CPU market by 2030, while global data-center electricity consumption could more than double.In September, McKinsey & Company warned that , with task expenses varying by as much as 30 times.About 60% of surveyed organizations planned to increase AI spending, while Levie cited research suggesting cheaper models could reduce token costs by 15 times.Earlier this year, OpenAI board chair Bret Taylor said , dashboards and web forms, but companies still face challenges in building, maintaining and preventing errors in these systems.Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Read Also: Photo courtesy: Shutterstock