The optimal design for a 'jet fighter without a pilot' rapidly converges on a missile.

Does it though? I suppose it heavily depends on your definition of "fighter", but cranked-kite and flying wing designs have been the go-to for subsonic stealth jets and UCAVs. Under a "something that is intended to kill other planes" definition a subsonic stealth air-to-air platform with good sensors might qualify as a fighter, but many people seem to think "fast, hypermaneuverable" are requirements too. Confusing the goal with the method, IMO.

Another major design constraint difference between "unmanned fighter" and "cruise missile" would be the deployment and reusability goals. Cruise missiles tend to need to fit in launch tubes or under wings, while presumably an unmanned fighter would not be so constrained (and thus be able to use larger wings). And while cruise missiles are very single use, the sensors on an unmanned fighter may make it too expensive to discard that way.

Jamming, hacking, possibility of being exploited(thing adversarial ML) etc etc

If we get real self driving cars how long till someone figures what you need to plaster on a tree to fool ML into crashing a car :) now imagine full nation state behind such effort. The risk of not having human in the loop is just too great.

You also just can’t do anything limited. I mean you can have even now fighter jet just kill anything it finds that does not transmit id, good in war maybe not so good at other times. Aegis is old tech and it is automated to kill anything that looks like a missile in full auto mode.

Iron dome probably same thing, engages and fires on full auto. Now imagine you hook it up to some automated 155mm battery, the moment missile get painted, it would start counter battery, I am guessing the round would land in Gaza before the rocket hits Israel. Now how many people would be ok with this is another question.

I wonder if the jamming problem could be solved by having a satellite relay communication to the jet via a laser beam, assuming the receiver on the jet can be tracked precisely. Perhaps also having redundant communication relay links from nearby jets. That would be harder to jam, right? Adversarial attacks could happen, but given enough diversity in sensors and algorithms I think it could be made reasonably robust.

I mean, the benefits of a pilotless fighter jet are pretty significant: more maneuverable, faster, easier to design, and probably cheaper. I'm especially curious about fighter pilots' opinions about this.

We already have a fairly good auto pilot. The complexity lies in rules of engagement.

Can you give an example of such complexity?