Reblogged:Two Sides of the AI Coin

[Gus Van Horn blog]

I have recently come across commentary about artificial intelligence from three very different sources. Of the three, the first one, chronologically, offers us the following still-valid summary. Mathematician Richard Hamming, lecturing at Naval Postgraduate School noted:

[T]ake artificial intelligence. The predictions made by almost all the experts 10, 20, 30 years ago have not been realized.

This he said in 1995. Perhaps some of the predictions he was thinking of have come to pass by now, but Hamming's point was that attempting to predict the future state of one's field can be extremely difficult.

In any event, practically everyone sees AI as potentially game-changing, but it is quite tricky to work with, as medicinal chemist Derek Lowe recently observed regarding its potential in drug discovery:

It fails as inspiration and as humor, but it makes a good caption for this photo: "Our view on modern technology is destructive." -- Inspirobot (Image by Sarah Stierch, via Wikimedia Commons, license.)

Even if you're not a computational drug discovery type, a look at Pat's roundup might be instructive, because seeing the actual problems that the field is wrestling with will very quickly take the shine off a lot of hyped-up headlines and press releases. These include things like "How do we even estimate the uncertainty in our model, and how do we compare it to others?", "How do we deal with molecules as three-dimensional objects with changing conformations, as opposed to two-dimensional graph-theory objects or one-dimensional text strings?", "Since no one can actually dock a billion virtual molecules into a protein target, how can we reduce the problem to something theoretically manageable without throwing away the answers we want? And how will we know if we have?" and "What do we do when our model will only start to work if we feed it more data than we're ever going to have?" The next time you see a proclamation that everything's been made obsolete by AI-driven modeling, keep those in mind. [bold added]

A third -- and somewhat surprising -- source (HT: GeekPress) illustrates Hamming's and Lowe's point quite humorously: It's about a gazillion inspirational posters generated by an AI called Inspirobot that "often turned out to be hilariously dark."

After seeing enough of these and being unable to escape the HAL 9000-like glare of the site itself, I wondered if the whole thing was really a gag. But some of what it generated could have been inspiring or even astute coming from a human, such as, "Our view on modern technology is destructive," but perhaps with a more appropriate image. Either way, the AI isn't foolproof, be it to craft inspiration or humor. It might be tempting to add, "even for a relatively simple task like making an inspirational poster or a joke." But to do that would be just as much to underestimate the human as we are tempted to overestimate AI.

-- CAV

Link to Original