Gary Marcus Questions AI’s Business Readiness Amid Optimistic Forecasts

In a recent interview, AI skeptic Gary Marcus—an emeritus professor of cognitive psychology at NYU and long‑time commentator on artificial intelligence—warned that today’s AI systems still fall short of the promise many hype cycles suggest. Marcus highlighted four core limitations: a lack of common‑sense reasoning, persistent unreliability, difficulty generalizing beyond training data, and challenges with compositional understanding. He also noted the steep cost of cutting‑edge models, raising doubts about whether AI will deliver broad business value outside a handful of niche applications.

Contrasting Marcus’s caution, OpenAI CEO Sam Altman has declared that artificial general intelligence (AGI) is essentially solved and predicts that by 2025 AI agents could enter the workforce and materially alter company output. The author’s conversations with business leaders using AI at scale suggest a middle ground: significant value can be extracted, but it requires substantial human engineering, customization, and expense.

One such leader is Euro Beinat, executive vice president and global head of AI and data at Prosus, a Dutch investment firm with a diverse portfolio spanning food delivery, e‑commerce, and fintech. Beinat’s experience across varied sectors provides insight into how AI is adopted in different business contexts. According to Beinat, while AI can drive gains, achieving reliable results often demands tailored solutions and extensive human oversight—far from the plug‑and‑play approach implied by AGI narratives.

The conversation underscores that, although the hype around AGI is intense, practical deployment of AI in business remains a complex, resource‑intensive endeavor that balances optimism with realistic expectations of the technology’s current capabilities.