It is true that a computer will take inputs and compute outputs, and that is all it ever can do. However, this statement does not quite end up explaining Space Invaders. There's a couple of steps in between that might have been glossed over slightly. [1]
A computer is entirely capable of generating novel (enough) output. Elsewhere I point to the trivial-seeming python function math.random(). Good luck predicting what that will do!
In reality, it turns out that math.random() and the theory behind PRNGs is not so trivial at all, and is actually quite interesting to dive in to. [2]
Your argument boils down to assuming there is true randomness in the universe. You need hardware level rng with a natural source of entropy to generate non-predictable random numbers, but even then you're just at the point where maybe we could predict the randomness if we could go low enough.
We can't, so at the moment it looks as though the universe might have true randomness (I'm very sceptical of this).
But even then, so what? Does that mean free will is just decisions with some true universal randomness thrown in? How does that give you free will if it is just random? And then we're back to, what does free will even mean.
Well, funnily enough rand() isn’t all that random, but that’s beyond the point. I don’t think randomness is a good argument for free will, that’s just nature. Does a flipping coin have free will? An electron before it decoheres?
Well, randomness is definitely novel output though.
PRNGs are not-quite-random, they're actually chaotic functions.
Which is actually a bit closer to how I think free will functions if you look at it empirically from the outside. (At least: if you were to use your <free will> to try to generate 'random' numbers; which will also not be perfectly random!)
Of course, it does sort of depend on your definition of <free will>.
Oh shoot, I see why I made the error, but on checking more sources to back me up, it turns out only some PRNGs are considered chaotic, and the most used ones are not necessarily viewed with that lens in literature. Eh, it would have been easier. In future I'll have to switch to a different set of algorithms to make my argument,.
A computer is entirely capable of generating novel (enough) output. Elsewhere I point to the trivial-seeming python function math.random(). Good luck predicting what that will do!
In reality, it turns out that math.random() and the theory behind PRNGs is not so trivial at all, and is actually quite interesting to dive in to. [2]
[1] This particular argument inspired by https://journals.plos.org/ploscompbiol/article?id=10.1371/jo... . It mentions using neuroscience to try to understand space invaders, which I think is funny. :-)
[2] https://en.wikipedia.org/wiki/Pseudorandom_number_generator , or eg. https://en.wikipedia.org/wiki/Logistic_map being used as an early PRNG