Companies may be betting on a world where they replace humans with artificial intelligence (AI), but that bid is failing 95% of the time, according to MIT’s NANDA Program. OpenAI’s much hyped GPT-5 model, hyped to gain ‘PhD-level intelligence’, was so underwhelming new model that paying subscribers revolted to have an older GPT-4o model back. At the same time, CEO Sam Altman went about changing the definition and goalposts, for the very foundation of company’s existence — a still elusive artificial general intelligence, or AGI.

Apple’s AI deficit may seem most visible one to thread addicts on X, but look beyond the hype, and you’ll realise AI models cannot even paraphrase information from their training sets with any sort of reliability. The problem is partly technology, which I just touched upon, and partly businesses across verticals over-estimating what an AI can do within their workflow. The fact that a ChatGPT or a Grok may work well for you in personal usage, doesn’t mean it’ll understand organisational nuances including legacy systems, real-world workflows and most importantly for the accountant, whether AI is translating into a profit.

Last time on Tech Tonic:

© hindustantimes