Did an OpenAI cofounder just pop the AI bubble? ‘The models are not there’

Andrej Karpathy, a founding member of OpenAI, has significantly tempered expectations for Artificial General Intelligence (AGI), asserting it is at least a decade away, contrary to more bullish industry forecasts from figures like Sam Altman and Elon Musk. He criticizes the overhyping of current large language models and unreliable AI "agents," warning that their lack of robust reasoning and potential for security vulnerabilities could lead to widespread inefficiency. This sobering assessment suggests a more protracted and complex development path for advanced AI, potentially recalibrating investor expectations and impacting valuations across the rapidly expanding technology sector.

Analysis

Andrej Karpathy, a key figure in AI and OpenAI co-founder, has significantly tempered expectations for Artificial General Intelligence (AGI), stating it is at least a decade away. This contrasts sharply with more bullish forecasts from industry leaders like Sam Altman (AGI by 2030) and Elon Musk (AGI this year or next). Karpathy criticizes the industry for exaggerating current AI capabilities, particularly large language models (LLMs) and "agentic" systems, labeling much of the current progress as "slop."

He specifically highlighted the unreliability and brittleness of current AI "agents," which are pitched as autonomous digital workers. Karpathy argues these systems lack robust reasoning, have limited environmental perception, and struggle with tool usage, leading to unpredictable results. He warned that poorly executed AI agents could lead to "mountains of slop" and increased security vulnerabilities across software.

This sobering assessment, which carries a "strongly negative" sentiment and high market impact, suggests a more protracted and complex development path for advanced AI than widely perceived. While Karpathy maintains that the technical challenges are surmountable, his comments imply a necessary recalibration of investor expectations regarding the near-term commercialization and reliability of advanced AI applications. The divergence in outlook from figures like Nvidia CEO Jensen Huang (who sees 2025 as the "year of AI agents") underscores significant industry debate.

Key Decisions for Investors

• Investors should recalibrate near-term expectations for AGI and autonomous AI agents, considering Karpathy's decade-long timeline and warnings about current system unreliability.
• Prudence dictates a focus on companies demonstrating robust, proven AI applications or foundational infrastructure rather than speculative agent-based solutions.
• Monitor for concrete advancements in AI reasoning, long-horizon planning, and secure system design as key indicators of genuine progress.