AI in Physical Environments: Companies Struggle with Accuracy and Cost

Executives at Honeywell and Ecolab discussed the limitations of AI in physical environments during Fortune's AIQ Summit. Suresh Venkatarayalu, Chief Technology Officer at Honeywell Technologies, highlighted the problem of accuracy, stating that customers demand '99.9999%' precision from industrial systems, while frontier models typically achieve only 85%.

AJ Wijesinghe, Ecolab's Chief AI Officer, emphasized the cost issue when implementing high-performing AI models in production environments. He noted that the 'tokenomics' for these models can be prohibitively expensive and even more costly than employing humans. However, by optimizing their models, Ecolab was able to reduce token costs by 70-80%.

Venkatarayalu emphasized that the move towards AI is not towards full autonomy but rather a 'semi-autonomous world' where operators build trust in systems. Wijesinghe also stated that agent technology is not mature enough for widespread adoption, and instead, companies should adopt a 'human-in-the-loop' model.