The rise of specialised AI models

When ChatGPT first hit the market, it was all anybody could talk about. I mean, a chatbot that performed like a Google search and tended to agree with you? Who wouldn’t love that, right? Slowly, pop culture saw ChatGPT getting used like a verb, again, like Google. It has reached a point where many people often discuss their deepest grievances with ChatGPT. However, if you’ve been using LLMs often, you've probably noticed this: they are all different, and there’s no single one that can replace the rest. Basically, there’s no room for a monopoly.

ChatGPT is usually still the go-to model for broad, general-purpose and multimodal business tasks. It is still the go-to model for many tasks. Claude is known for its comprehension skills and nuanced analyses, especially when dealing with long documents. Then there is Perplexity, which is very good for research. Basically, each model has its pros and cons; if used in the right way, it will lead to the best outcomes. This also means that consumers need to get a better idea of LLMs and understand the nuances for optimum results.

This blog post suggests that users need to learn “task matching”, where the user understands the task at hand and matches it with the AI tool most suitable for it. This means one thing: the hiring market will shift. The rise of specialised LLMs does indicate that the need for domain experts will continue to rise. This would suggest that the hiring requirements would change from looking at a jack of all trades to finding someone who is best suited for a particular set of skills. The job of the generalist is not over, but the specialists are shinier now, in a sense.

This indicates that companies and startups will tend to develop AI models that are geared towards specific fields like law, medicine, and finance. The bigger models won’t disappear, though. Think of the future models like a stack, where the base layer would be the big model and multiple specialised AI tools (from engineering to legal) would be stacked on top of each other. The team would also reflect a similar version: one generalist working with many domain experts. It’s not bleak, just different. Not one for all but all for one.