To Practical Use

Dearest Rachel –

I honestly don’t know if I’m boring you with my latest obsession with artificial intelligence and pictures of you. Despite your degree in computer science (which you earned essentially pre-internet, so it was practically worthless within a few years of your graduation), you rarely seemed as interested in the ins and outs of the machines and what runs them as I have been (although what little I demonstrated to you when I was cleaning up your parents’ vinyl album collection seemed to impress you). Moreover, in the month or so that I’ve been at this project, I’ve treated it mostly as a toy, to put you up on a pedestal, sometimes quite literally.

But as it turns out, there’s a practical application to all this artificial intelligence stuff. Actually, there are a whole host of applications – how would research into the field ever get off the ground without time and money poured into it, and who would do this without expecting at least some kind of payoff? Many of them, of course, are still in the theoretical and experimental stages (and one that I’m planning to show you is actually still in ‘beta’ development, even), but it’s kind of amazing to see where we are and how far we’ve come in just a couple of years, or even a few months.

I mentioned that time and money is being invested in this field; it shouldn’t surprise you that one of those big-name investors is Microsoft. I’d never noticed this before, but in Excel they have an option on their main ribbon that allows you to ask the program to analyze your data for you: this is Microsoft utilizing AI to allow a layman to make sense of the numbers being generated, say, by an accounting program.

Now, from my perspective, it’s not showing anything earth-shattering; the scatter graph at the top right, showing a general correlation between actual and budget figures, is what one would expect to see when one compares the two sets of data (since the idea is to predict what one’s expenses ought to be and stick with in reality), as is the lower one showing a similar correlation between last year’s activity and this year’s. But it’s a visualization I never would have thought of assembling, even as it confirms my expectations – even the ‘outlier’ indicated on the upper chart still verifies the correlation between the budget and actual figures (it’s just a particularly large category of expense, in both cases) – and the computer, through the AI, offers dozens of similar breakdowns of the data in ways that it thinks I, as the user, might consider informative, allowing me to look into their menu of options as I wish.

Admittedly, it’s practical to the point of almost being boring (unless you’re into that sort of thing). In which case, let me show you something I discovered just yesterday that actually meshes well with the work I’ve been doing, particularly with my attempts to assemble your parents’ old photos. It seems that Adobe is beginning to incorporate the technology into their programs as well, in this case, Photoshop. I honestly couldn’t recall if I had started assembling our collection of pictures together in a single massive folder before you left us…

…but most of the ones I’d assembled up until recently were digital pictures. The ones from your parents, on the other hand, were all on photographic film, and as a result, relatively low quality (although, in fairness, not necessarily bad for their day). Compared to what we’re used to seeing, those shots from around the turn of the millennium are pretty grainy.

It so happens that Photoshop has begun work on something called Neural Filters, which can be applied to an old photograph like this one.

Anyway, the results rather speak for themselves:

Granted, the computer seems to have stolen the buttons off of your dad’s sweater, and his face seems to have lot a little clarity (possibly from being on the very edge of the shot, to the point of being cut off), but your face and Daniel’s really get sharpened up by the process. It’s a process that, with a little practice and marketing, could be a nice side hustle if I really wanted to do that with myself; I’m sure there are plenty of people who would like their old photos restored in this way. Not sure that I’m necessarily interested; even going through the two hundred or so pictures I’ve scanned thus far would take weeks, if not months, to upgrade, while my current project offers rather instant gratification. Still, it’s interesting to see a more practical application to this little hobby of mine.

To be fair, however, it’s not a perfect system yet. As you can see, it’s still considered to be in its beta stage, and some things, it just can’t recognize. This picture of you by your original PT Cruiser, for instance…

…utterly confused it, as it couldn’t seem to recognize your face as a face. As a result, your exultant expression, obscured as it was by the bright sunlight, couldn’t be enhanced by the neural network; at least, not yet. So, not every picture is capable of ‘improvement.’ Then again, maybe I’m just not making the right adjustments. Who knows? Obviously, there’s a lot of room for learning, both on my part and the computer.

And with that being said, keep an eye on me, honey, and wish me luck. I’m gonna need it.