Slow, yes, but brains are 5-6 orders of magnitude more efficient than present-day silicon. In any case that's not the point. By acknowledging you've already given up the argument of AI doomers.
The Bostrom-Yudkowsky argument for existential AI risk is based on one of two things being true, both of which they have argued in the past: (1) AI minds can be of qualitatively different order and power than human intelligence, or (2) recursive self-improvement will make them transcend on radically short timescales. Actual events have shown both to be false.
Famously there was a large debate between Yud and Robin Hanson about this issue, with Yud basing his predictions on us going from AGI (e.g. ChatGPT in 2022/2023) to ASI in mere hours, and therefore kill all humans in no time at all, and Hanson arguing that it would be a slow takeoff, and the incentives for slow takeoff make cooperation and coexistence vastly more likely.
We are without any doubt squarely in the Hanson timeline, and the Bostrom-Yudkowsky super intelligence theory debunked by reality. Yet very few AI x-riskers seem cognizant of this, and fewer still have actually updated their priors.