Faith in God-like LLMs is waning

When tech folk talk about the lacklustre progress of large language models (LLMs), they often draw an analogy with smartphones.

- 

The early days of OpenAI's ChatGPT were as revolutionary as the launch of Apple's iPhone in 2007. But advances on the frontier of artificial intelligence (AI) have started to look akin to ho-hum phone upgrades rather than genuine breakthroughs. GPT-5, OpenAI's latest model, is a case in point. It has generated even less buzz than is expected from the iPhone 17, Apple's newest release, which is due to be unveiled on September 9th.

The slowing pace of improvement at the bleeding edge of generative AI is one sign that LLMs are not living up to their hype. Arguably a more important indication is the rise of smaller, nimbler alternatives, which are finding favour in the corporate world. Many firms prefer bespoke models that they can tailor to their specific needs. These so-called small language models (SLMs) are cheaper than all-purpose LLMs, whose God-like intelligence can appear superfluous. As David Cox, head of research on AI models at IBM, a tech company, puts it: "Your HR chatbot doesn't need to know advanced physics."

Besides being as easy to run on a company's in-house IT systems as via a cloud-service provider, SLMs may also be more useful for AI agents, which do work-related tasks alongside or instead of humans. SLMs' slimmer size makes them particularly well suited for AI in smartphones, self-driving cars, robots and other devices where energy efficiency and speed are at a premium. If they continue to become more reliable, they could validate the decision of Apple and other device-makers not to follow the herd by investing gazillions in cloud-based LLMs.