I think there are a few important factors to consider (I did not go through all the comments btw so apologies in case, there are too many). Historically lower development costs have also expanded the scope and ambition of what gets built. With LLMs acting as assistants and humans focusing more on architecture and decision-making, teams can explore multiple solutions in parallel and tackle projects that were previously too expensive or complex. That may increase, rather than decrease, the need for people who can manage complex domainy.
There are also practical constraints that could slow or reshape this trend: the growing cost of AI infrastructure and data centers, environmental impact, regulation around copyright and training data, and a likely shift toward specialized domain-specific models instead of a few generic ones. On top of that, if LLMs become the new search engine, advertising and commercial incentives may erode their neutrality.
Finally, LLMs depend on a constant supply of high-quality human-generated data, but they are also reducing the production of that data (e.g., the collapse in Stack Overflow activity). We are also in a very unusual market phase, with a handful of companies heavily investing in and cross-subsidizing each other. It is not obvious that the current pace of scaling can continue indefinitely, or that today's AI landscape represents a stable long-term equilibrium.
There are other factors, like adoption of frontiers LLM in the market that is anyway slow, also the impact of AI in the economy that is anyway based on humans that must be controlled (because it has consequences when it comes pensions, consumptions of goods, etc.. )
Sorry: I used LLM to summarise my points to avoid long wall of text :D
There are also practical constraints that could slow or reshape this trend: the growing cost of AI infrastructure and data centers, environmental impact, regulation around copyright and training data, and a likely shift toward specialized domain-specific models instead of a few generic ones. On top of that, if LLMs become the new search engine, advertising and commercial incentives may erode their neutrality.
Finally, LLMs depend on a constant supply of high-quality human-generated data, but they are also reducing the production of that data (e.g., the collapse in Stack Overflow activity). We are also in a very unusual market phase, with a handful of companies heavily investing in and cross-subsidizing each other. It is not obvious that the current pace of scaling can continue indefinitely, or that today's AI landscape represents a stable long-term equilibrium.
There are other factors, like adoption of frontiers LLM in the market that is anyway slow, also the impact of AI in the economy that is anyway based on humans that must be controlled (because it has consequences when it comes pensions, consumptions of goods, etc.. )
Sorry: I used LLM to summarise my points to avoid long wall of text :D