Elsewhere: Lovecraft As AGI Prophet, Community Updates

I'm now on Notes and in Discord

🦑 Regular readers have probably spotted Lovecraftian horror as a long-running theme in my posts. Among the very first images I ever generated with generative AI was a set of Midjourney beta generations of Cthulhu rising up from the ocean:

I’ve since sprinkled little bits of Lovecraftia (is that a word?) in my posts here and there, but I always felt like a bit of a poseur because I’d never actually gotten around to reading the man himself. All my Lovecraft lore came from RPGs and board games. At least, that was true until this past Halloween, when I finally sat down with a big omnibus edition of his complete works.

Actually reading Lovecraft himself had a weird effect on me, which I told some friends about but hadn’t otherwise written about publicly… until yesterday in

RETURN

:

H. P. Lovecraft, Prophet Of AGI

Hard to know what to excerpt from this, because I pull out a number of contact points between Lovecraft and the present AI/algos/AGI discourse. I end up thinking about the issue of safety, not so much physical but spiritual and psychological:

But the alignment-related catastrophe that worries me is not the X-risk scenario where an AGI murders us all. Rather, I’m worried about an AGI that makes us not want to live anymore — an AGI that brings on a spiritual catastrophe by convincing us to simply give up because what is the point?

An astute RETURN reader and fellow substacker (writing at

A Grain of Paprika

) actually posted a great comment on the piece, drawing an analogy to retirement. Retirement has weird (and sometimes deadly) effects on people. What happens when humanity essentially retires?

---

🧷 I also published a piece in City Journal on the topic of “AI safety,” making the point that this entire discussion is greatly hindered by the fact that, as a society, we’re currently contesting the definitions of both “intelligence” and “safety.”

The AI safety debate couldn’t have arrived at a worse time in our history. Both machine-learning researchers and our larger society are bitterly divided over what two of the discussion’s key terms—“intelligence” and “safety”—actually mean.

America’s post–George Floyd era “racial reckoning” has seen a rapid public rethink of what intelligence is and how it should or should not be measured. Colleges and professional schools are ditching standardized tests under pressure from equity advocates, who insist that these tests are slanted toward a narrow, racialized conception of intellectual competence that unfairly discounts what nonwhites have to offer universities and professional guilds.

But it’s not just our broader society that’s divided over the nature and meaning of intelligence. Researchers can’t agree on a rough working definition of this elusive concept to measure properly if or how their increasingly sophisticated machine-learning models are exhibiting more of it. Machine-learning experts offer competing definitions of “intelligence,” along with a variety of benchmarks for assessing it. Market leader OpenAI has its own, more practical definition of “artificial general intelligence”—“highly autonomous systems that outperform humans at most economically valuable work”—but even this is slippery enough to be contested.

I end up coming out against efforts to slow or pause progress in AI, which is no surprise to readers of this newsletter. I just can’t see unilaterally disarming in what is essentially a new arms race while we sort out the definitions of the words we’re using to fight each other over the role this new tech should play in our society. As I sometimes say on the bird site: don’t hate, accelerate.

Community update

Substack has launched its Notes product, and I’ve started posting there on occasion. The following note, for instance, is a tangent I cut out of the Lovecraft piece linked above:

I also have a subscriber chat enabled on Substack, but it’s not that active. Frankly, I’m not a fan of Substack’s on-site chat. There’s nothing wrong with it, per se, other than it’s just not Discord.

I’m currently using Discord for community and will continue to do so. In fact, I’m actually just now really ramping that up and will finally introduce the long-promised “paid subscriber”-only channels, soon.

You can log into my Discord here and go through the onboarding flow I’ve set up. It collects your email address so that in the future I can check it against your subscriber status (once I write the code for this) and give you the correct role.

My hope for the community is that it’ll be a resource for myself and others to stay current on up-to-the-minute AI news (acting as a filter for the constant firehose of product announcements, papers, and threads) and that it’ll also be a place where builders can gather and learn from one another.