AI Models and Agents Will Specialize Over Time, Just as The Software Industry Did
While I do believe that Large Language Models are commoditizing, my view is that outside of the large flagship models built by Google Gemini, OpenAI and Anthropic, there will be thousands of smaller models - language and multi-modal - that specialize in specific disciplines and functional areas much like software applications have proliferated and evolved since Microsoft’s founding in 1977 out of Alburquerque.
AI agents and models will find natural niches where they may operate most profitably. AI agents and models will find specialization much like water finds the lowest point. Highly specialized models will expertly perform specific functions optimized for specific use cases much like humans are highly specialized across disciplines and functions.
While companies such as OpenAI, Microsoft, Google and AWS scratch the surface on vertical use cases, I doubt those companies will create highly specialized models, AI agents and AI-infused applications for vertical-specific use cases. Rather, I expect that entrepreneurs from industry will leave those careers behind to start technology companies that build models, AI agents and AI-infused software applications that address acute business problems that founders will be intimately familiar with. This is exactly how the Software industry evolved.