The real question to ask is;

Is it still worth pursuing a software startup ... using the same strategies/approach/mindset/tools/means/methods of yesteryears?

Answer is Yes for the first part and No for the second part.

Problems which need software solutions and/or have a software component (at this point, practically everything) still exist in plenty. AI/ML has only changed the playing field drastically. It cannot understand the intent behind the problem which is a uniquely Human Dilemma. This is where the opportunities lie; understanding different Problem/Business Domains by studying and closely interacting with possible clients and coming up with tailored solutions whether domain-specific or client-specific.

The emphasis now is on better problem understanding/specification/verification (DSLs/notations/diagrams etc.) and faster iteration to a MVP. AI/ML is a great help here but the cycle is initiated by a Human who is always in the loop and steers it on the right trajectory in the state space of possible solutions.