CronoDAS comments on Current AIs seem pretty misaligned to me

[CronoDAS]

Is a scenario in which AI “usefulness” bottlenecks on alignment instead of capabilities before it can become seriously dangerous an optimistic one?