Machine Learning: You can’t teach an old dog new tricks ...
… but you can teach a machine anything. Indeed, gentle, well-inclined readers: Sometimes, even answers to complex questions are simple. The era of Artificial Intelligence has only just begun, the soft silver shimmering light of the new dawning age is just climbing over the horizon of humanity.
By Strigalt von Entf*
And yet, it is already apparent that the newest creation of Homo Sapiens will reach, overcome, even reduce to dust all thinkable barriers on the double. Already now, there are AI that diagnose more precisely than the most experienced doctors, and DeepBach composes chorales that would bring blank awe to the wig-clad face of the baroque virtuoso who gave the AI its name. And let us not forget the presumptuous computer that taught itself to defeat humans at Go, a game which the common Central European had not even known until they heard that AI is now better at it. From this view, the beginners' course is hardly worth it anymore. What will become of us humans when AIs have taken over everything for the common good? What remains is the hope that a clever machine will create a matrix for us humans in which we can dream of a world without AIs, where we are still useful.
What is interesting here is that the artificial systems only start becoming really good when we, their creators, no longer understand how exactly they actually learn. This game of confusion is taken to the extreme by the fact that AI is now also taking over the development of new AI. We are left standing by in amazement and watching the golem come to life, soon to elude our introspection. I wonder if God felt the same way. Is that why he now discreetly stays out of our lives, refraining from appearing in burning bushes and sending downpours of frogs, because he is henceforth only able to stand next to his creation with a sceptical gaze, one question circling in his only partially omniscient head: "What the hell are they doing?"
This game of confusion is taken to the extreme by the fact that AI is now also taking over the development of new AI. We are left standing by in amazement and watching the golem come to life, soon to elude our introspection. I wonder if God felt the same way.
Most revered fellow creatures, we could, of course, spend our time dwelling on the big questions of humanity and exploring the potential impact that AI have on them, but, as many others are already doing so elsewhere, our limited space is in fact too precious for that. Therefore, let us instead ask the one crucial, ground-breaking question: Can an AI learn to write a better column than the honourable Strigalt von Entf? Don't be hasty in your answer! I pulled a few strings and set the unthinkable in motion. Based on my past insights into the world of technology, which you are surely well aware of, a specially created AI named "deepgalt" has been trained to write a well-honed, deeply thought-through, stirring and refreshing column. I hereby leave you and myself in the hands of the digital scribbler.
Esteemed people who read letters,
oh, what a lofty tragedy is about to unfold and confronts us with perplexity: what does man do, where does Homo erectus, who has lost the key to his paradise. Deep Learning. Deeper Learning. What follows deepest learning? Shallow learning. The shoals that Ulysses had to navigate around, peppered with Scylla and Charybdis, monsters, monstrosities, mother-in-law's kidney stones.
Artificially intelligent is he who is not artificially stupid, or not naturally intelligent, or of course both. What about artificial stupidity. Various applications for artificial stupidity are conceivable: Funny search engines, SPAM emails, how to manipulate the latest captcha.
Stupidity is productive, intelligence remains in brooding. MacBeth, the doer, he did. Max and Moritz, team, complementary, yin and yang of the modern-age fairy tale. Then at the end they end up in the flour box. Max: SPAM. Moritz: SPAM!
Well, dear readers ... do you think what I think? I am sure you do. What are we looking at here? A melange of borrowings from ancient mythology, anthropological references to the evolution and the essence of humanity, everyday practical application and lifeworld references: I believe we have to admit that even after only a few learning sessions, "deepgalt" managed to form some really profound thoughts. I'd be lying if I said that didn't do anything to me. Yes, that's how poor Christian of Neuvillette must have felt when he realised that long-nosed Cyrano was forging much better words to sweet-talk his beloved Roxane.
I don't have to tell you how he fared ... but well ... we're not there yet, no, dear readers, I'm still in the saddle, still holding the reins, i.e. the keyboard. Strigalt von Entf has not yet served his time. And neither have you. Someday, ladies and gentlemen, the machines may do everything better. Until then, we are left with the realisation: Deep Learning still has a lot to learn. We can and have to enjoy the time, during which we are still good for something.
Your Strigalt von Entf
*Our "Feuill-IT-ong" format is created in collaboration with the two freelance writers Tobias Lauterbach und Daniel Al-Kabbani who occasionally contribute to the satire platform "Der Postillon". Under the pseudonym Strigalt von Entf, they report on current events from the world of technology – always with a wink! ;-)