The interpolation would tend to at best half a pixel? And the phosphor lag (like on a tube) would be an issue surely?
Are there instances of single eye outcome where the subject has drawn perceived image so we can understand how this exposes into conscious visual stimuli?
Even just a flash on the left == left object vs flash on the right == right object would be a useful signal compared to zero. But describing it as "vision" would be stretching it. 378 pixels is a few letters at 10x18 so it's 2-3 words. Again, massive gains on nothing, but it's beyond "large print" its "large print with a magnifying glass" and it might be phosphor burn colour against black or a foggy field, or a number of things.
To be clear, this is amazing stuff and hats off to anyone who helped make it happen, but let's not assume we're in "snow crash" territory just yet.
The lag would be in signal processing external to the user.
Interpolation would be more transparent, much like it is for you right now. There are no phosphors in tubes in any of this.
I made no such "snow crash" assumption.
Users of devices like this have described their experiences and those are not generally big square pixels.
Think of those more like points the brain can do something with.
The chip stimulates the remaining neuro-signal entities present in the damaged retina. I doubt there is a 1:1 relationship between those and the signaling points on the chip.
When the company can do better than on/off bright/contrast, the overall experience should improve dramatically. There will be more signal points (1024 ish?) and those having variable output levels will give the users visual cortex a whole lot more to work with.
About the only analogous thing I can come up with is cochlear implants. Those have a number of signal points that seems a lot smaller in number than expected. That was certainly my take. The more of those there are, the more concurrent sounds can be differentiated. A greater sense of timbre, in other words, becomes possible.