Lovecraft, Elon Musk & Superhuman-AI

NOTE: This is an archival backup, original is on medium.com, please read it there.

It’s hard to look at H.P. Lovecraft (1890–1937) as a pioneer in warning people of the dangers of AI. And even more bizarre to see him in the same boat as someone like Elon Musk.

I mean, yeah, “cosmic horrors” can easily painted in modern sci-fi as originating from “perverted” superhuman artificial intelligence. But Lovecraft had neither interest in modern information technology (that wasn’t yet invented, btw), nor much interest in science and technology based fiction in general. He was no Jules Verne.

Nevertheless, by means one cannot even begin to fathom, a strange meme has landed in his darkly creative (and xenophobic, and in other ways disturbed too…) mind. A strange intuition of things yet to come:

Shoggoths — described as “a shapeless congeries of protoplasmic bubbles” with “myriads of temporary eyes forming and un-forming”. Their image seemingly evocative of a giant amoeba-like biological entity:

And there are two things that make them very special:

• their function: they were initially “used to build the cities of their masters” — they were in fact self-assembling, self-replicating robots (which he fashioned as being made from of biological instead of electronic components, obviously a more plausible variant for Lovecraft given the state of technology at his time)
• their intelligence: absent at first, when they were controlled by “hypnotic suggestion” (had he lived in modern times, Lovecraft would’ve been more inclined to use the term “(software) programming”!)… but later developed intelligent minds and rebelled

And there’s a third thing which also makes their story sound too familiar: their masters became dependent of them. Although the first Shoggoth insurrection was quelched, they were too crucial for the Elder Things’ existence, so their “usage” continued despite their danger.

Now wind forward the time like 90 years or so, and take a look at today’s worries about… Artificial Intelligence. The core of this pompous term is actually the engineering of something called “machine learning algorithms”. It turns out that crafting intelligent machines (at software or hardware level, doesn’t matter), is really, really hard if you try to program them to be intelligent. One can instead program them to learn from data. And once you do this, there’s no reason why you can’t get to a machine that can learn to improve itself. That’s a severely underwhelming summary of the present and future of modern “AI” and/or “Machine Learning”. But it’s good enough to support this… rant.

Also, our dependence on AI: it will surely only grow and grow…

And, more worryingly, there’s good reason to believe that what we call “self-consciousness” or “free will” are nothing but emergent phenomena that will arise once a large number of powerful and general AI agents will start interacting.

One could argue that we can avoid designing systems prone to “self-consciousness” (makes it sounds like a disease… and maybe it is, maybe we are “sick” with it!), and also build solid fail-safe mechanism and “preprogrammed morality” into our AIs.

Many bright human minds have embraced this point of view. Elon Musk, who I mentioned at the start of the article, is one the most vocal pioneers of the field of “AI safety”… and he’s also the one that convinced people this is no joke! (He even used the words “summoning the Demon” — scary words to hear from a reputable engineer and billionaire businessman.) One could say it could all turn out to be a joke, just like promises of flying cars (and cured cancer)…

But the fact that some can be ignorant enough to believe it a joke, makes lending credibility to it using your public persona one of the most important things one can do. In an era where we’d all rather share funny memes on social media instead of thinking deep and working hard on the really important things, any extra spotlight matters! Heck, even Vladimir Putin recognized that AI will be a big deal — he’s in no way a “source of credibility” for anything, rather the opposite, but this shows you that shit got noteworthy!

Anyway… I will not go deeper that the eggshell of this. My purpose was just to point some of the missed connections.

But I will not refrain expressing my… pessimism. You see, there’s one thing that most are ignoring: the “automation of creativity” part, that’s getting hotter and hotter. Now, it may be that in order to create artificial intelligence capable of superhuman performance in any narrow domain, you can do it without any need for “self-consciousness” or “free will” (and, despite considering “free will” a misnomer myself, I think that what the words actually stand for, comes hand in hand with being self-conscious). And as you get to more general and general intelligence, and you demand more and more performance from it, you’re going to need “creativity”.

The thing with true creativity, is that, for reasons that I will explain in other articles, I’m fairly convinced that it requires both “free consciousness” and “the capacity for evil”. Sooner or later, the competitive advantage of one extra percent of creativity, will make some corporation or military reckless enough to let a few drops of these “poisons” fall into the genesis cauldron of their next AI agents, which will then come to evolve beyond human level. At that point, I have only hope that we’ll either have decent mind uploading capability so we can “join the Gods and be like them”. Or at least have the luck to be among the survivors that will be left to live in a “reservation” (hopefully quite Eden-like …if they’re not much into “animal cruelty”).

Now, coming back to Lovecraft and his horrors, which seem almost childish now compared to what could really be coming for us, his 1931 novella At the Mountains of Madness is a good piece if you want to read more about the Shoggoths, which are also mentioned in other stories (and have been “borrowed” by other writers and comic books artists). But you will be disappointed — he didn’t go much in detailing the story of the collapse of the civilizations that engineered the Shoggoths. They remained as lingering forgotten horrors, beneath the Antarctic ice, waiting maybe for something like global warming to wake them from their aeonic slumber…

At least we’ve blessed ourselves with these imaginary horrors, to keep our souls somewhat warm and comfy. For the real ones… they will be beyond our comprehension anyway.