Mark Cuban identifies inconsistent output as primary weakness of AI

Billionaire investor Mark Cuban has issued a stark warning to the corporate world, identifying artificial intelligence’s fundamental inability to provide consistent answers as its greatest weakness.

Sharing his insights on X, the Shark Tank personality explained that while traditional software follows rigid logic, large language models (LLMs) operate on probabilistic rules.

This statistical approach to predicting the next word inevitably leads to varying outcomes, even when the provided inputs remain identical.

"I'm coming to the conclusion that the biggest challenge for enterprise AI, and AI in general, as of now, is that it's still impossible to make sure that everyone gets the same answer to the same question every time," Cuban noted.

He argued that this inherent inconsistency creates a "massive liability" for organisations that require predictable results.

Furthermore, Cuban used this technical limitation to counter "AI doomer" narratives regarding autonomous superintelligence.

He asserted that because AI does not understand real-world consequences or the need for consistency, it cannot achieve true consciousness or conquer humanity.

The investor emphasised that these flaws make "domain knowledge more valuable by the second," as human judgement is required to challenge and verify AI outputs.

This perspective is supported by recent industry data; Google CEO Sundar Pichai revealed that while AI generates 25% of the firm's new code, every line is reviewed by engineers.

Similarly, Uber’s Dara Khosrowshahi noted that their 10% AI-generated code requires strict human approval.

As the technology matures, the ability of skilled professionals to provide oversight remains the critical safeguard against the unpredictable nature of modern machine learning.