I feel like nowadays people are overhyping AI. Sure, AI can do and achieve a lot of great things. Sure, while it takes AI a lot of tokens to complete the results, the inference cost of such tokens keeps decreasing every day.

But I still feel like nowadays people are increasingly automating or replacing workflows, stuff, or workers with AI just because they want to replace them, instead of because the act of replacing them has positive economic value (ROI) right now.

Sure, I agree that doing such an act in the future would be flawless, since we assume the inference cost will keep dropping, which I don’t disagree with. However, I feel like we are too aggressive about using AI to replace existing tasks, or, more appropriately, we are over-adopting AI.

We are increasingly building infrastructure that we might not need as much as we think we will in the future. We are also funding startups and giving them more valuation than the value of the problems they solve, at least right now. In the future, due to long-tail theory, those companies might be able to survive and solve such niche problems at low cost because of technological advancements.

That’s what I wanted to say.

I have never felt such a strong feeling about AI.

I feel like, sure, AI can replace a lot of programmers’ manual work. But is the replacement itself reasonable, valuable, mandatory, or meaningful?

In the future, I would confidently say yes. Right now? I remain skeptical.