Instead, he envisions that the end result will be “a multiplicity of models that are more optimized, specific, and that address different problems.”
It’s certainly important to note that their company is focused on becoming a GitHub-like repo for exactly those types of specialized models, including both larger models offered by companies like OpenAI and Meta (e.g. gpt-oss and Llama 3.2) and fine-tuned variants adapted to specific needs by developers or smaller models developed by researchers. That’s basically what Hugging Face is all about.
So yes, it’s natural that Delangu would say this. However, he is not alone. In one example, research firm Gartner predicted in April that “the diversity of tasks in business workflows and the need for greater accuracy is driving a shift toward fine-tuned specialized models based on specific tasks or domain data.”
Regardless of which direction LLM-based applications go, investment in other applications of AI is just beginning. Earlier this week, it was revealed that former Amazon CEO Jeff Bezos will be co-CEO of a new AI startup focused on applications of machine learning in engineering and manufacturing — and the startup has launched with more than $6 billion in funding.
That could also be a bubble. But despite some of DeLongue’s statements on the AI bubble discourse clearly meant to promote face-hugging, there’s a useful reminder: The broader term “AI” is much bigger than just big language models, and we’re still in the early days of seeing where these methods will take us.