Superla-Pope Peeps

Eliezer Yudkowsky, Co-Founder and Research Fellow of an outfit called the Singularity Institute for Artificial Intelligence (SIAI) has recommended a short text of his posted on the blog Overcoming Bias as his response to my exchanges with Aleksei Riikonen and others in the Singularitarian Agony posts from the last couple of days (just scroll down a bit for these if you like). I responded to Eliezer's recommended essaylet on a discussion list we occasionally nip in and out of and then he and I sniped at one another very entertainingly a few circuits around the roller-rink. I regret that I can't re-post all that edifying snark -- since it seems to me content posted to that list is probably offered up with the expectation of a certain privacy I will respect -- but since the original blog-post Yudkowsky recommended as the corrective to my position is freely available I do think I can let you in on my response to that.

In Yudkowsky's text he proclaims that "what scares me is wondering how many people, especially in the media, understand science only as a literary genre."

Since he is offering up this thesis as one that applies to me, I can only assume he wants to paint me as a person who "understand[s] science only as a literary genre." Think carefully about what it means that Yudkowsky either can't see why he is wrong here, or why he thinks it is a good idea to peddle such a false impression.

Certainly I will attest that there are key aspects of technoscientific change that are well describable in cultural, social, and political terms. Because my training and temperament suits me to talk about just those aspects, and because I address my arguments to technocentric folks who often underestimate just those aspects to the cost of sense it makes enormous sense that those are the aspects I devote much of my attention to in my writing. But it is quite a leap from here to the conclusion that I think of science "only as a literary genre." I certainly don't believe any such thing. I challenge Yudkowsky to unearth such a reductive claim in my writing. One place to look might be my essaylet Is Science Democratic? Another, Technoethical Pluralism.

I daresay Yudkowsky should be able grasp, at least in principle, that even a scholar whose training best suits him to discuss the literary dimensions of the texts in which scientists seek to communicate their findings, their inspiration, their significance, and so on will still be able to distinguish those dimensions from the proper scientific dimensions (the collective practices of invention, testing, debate, and publication through which descriptions are offered up as candidates for warranted belief in matters of prediction and control and which satisfy the criteria hacked over millenia to select among these candidates, testability, coherence, concision, saving the phenomena, and so on) also exhibited in these texts.

Later in his post Yudkowsky offers up this bit of homespun wisdom:

"Probably an actual majority of the people who believe in evolution use the phrase 'because of evolution' because they want to be part of the scientific in-crowd -- belief as scientific attire, like wearing a lab coat. If the scientific in-crowd instead used the phrase 'because of intelligent design,' they would just as cheerfully use that instead -- it would make no difference to their anticipation-controllers."

I would be very interested to know more about the empirical experiences on the basis of which Yudkowsky has "induced" this "probability." I guess he imagines people in general too dull witted to affirm instrumental beliefs warranted by consensus science because these have proved especially good in facilitating their satisfaction of instrumental wants. Yudkowsky seems to think the poor benighted masses choose surgeons over faith healers so regularly only because they hope to be taken for the cool kids in lab coats in University Science Departments. It's an, uh, interesting theory.

I must say that the very notion of a "scientific in-crowd" so compelling in their allure that a dim-witted "Mob" parrots their utterances so as to be mistaken for them seems to me at once so flabbergastingly off-the-reservation and so baldly elitist that I laughed out loud upon reading it.

As an aside, I do agree that sometimes utterances that appear to affirm as warranted what are in fact absurd would-be instrumental descriptions ("An angry Patriarchal Creator-God with a gray beard sitting in a great Stone Chair exists," "the Earth was created in seven days," "my team won because Jesus likes us better than our opponents," and so on) are really functioning as social signals, indications of moral identification/ disidentification. And so I think, for example, that the worrisome reports one sometimes reads that vast numbers of Americans affirm idiotically Creationist beliefs in surveys are often actually using these affirmations of belief to signal to the surveyor and to an imagined audience something like the statement "I try to be a good person" -- where this correlates to membership in some church they haven't attended more than sporadically most of their adult lives. I suspect that misinterpretations of these reports often make sensible secularists and atheists panic unnecessarily about the state of this actually promisingly secular multicultural American society. But the proof is in the pudding: it seems to me that when people's circumstances demand actual instrumental beliefs their conduct is more pragmatic and scientifically informed than not, and this is scarcely because they want to be like idealized scientists. I think people in general are capable and sensible precisely to the extent that they have access to information and protection from duress. In other words, I really do think people are much more rational than they might seem to be when they report belief in superstitious nonsense and that they exhibit more rationality than they are often credited for in their actual instrumental conduct. It seems to me that many Singularitarians and Technocratic Elitists (perhaps Yudkowsky among them, given the above) have formed the opposite impressions, for whatever reasons, and to their cost.

Yudkowsky continues on:

"I encounter people who are quite willing to entertain the notion of dumber-than-human Artificial Intelligence, or even mildly smarter-than-human Artificial Intelligence. Introduce the notion of strongly superhuman Artificial Intelligence, and they'll suddenly decide it's 'pseudoscience.'"

I encounter people who are quite willing to entertain the notion that the streets are filled with actual bipeds, but who "suddenly" decide it's "pseudoscience" if you introduce the notion of perfectly bipedal angels into the discussion. Why, it's like a world gone mad!

You know, the world is littered with actually existing calculators and computers, many of which can already "outperform" normative exhibitions of human intelligence in some of its conventional registers. But, I hate to break it to Yudkowsky et al, there is nothing like an entitative post-biological superintelligence even remotely on the horizon, and so I think skeptics might be forgiven their "sudden" skepticism on this score in light of this alone. Surely it isn't exactly only "suddenly" that sensible people have suggested that worries about the imminent arrival of Robot Overlords might be a wee bit skewed and pseudo-scientific?

But, of course, things are much worse for our Singularitarians than that on my view: Not only are there common or garden variety computers everywhere but Robot Gods nowhere, there is also a long history of people who sound exactly like Eliezer Yudkowsky making fun of people like me (and much smarter and better informed than the likes of me) for doubting their hyperbolic pronouncements about the proximate arrival of world-transforming AI.

And, although you would never guess it from the withering contemptuousness Singularitarians sling at their skeptics, the AI guys are, so far, always, always wrong and the skeptics, so far, always, always right. While it's true that the endlessly reiterated failure of the Strong Program (let alone the, shall we say, Sooper-Strong Program Singularitarians have come somewhat perversely to prefer in the aftermath of these failures) doesn't earn anybody the deduction that the Program will always so fail (I don't hold that view myself, as it happens), one wonders why even caveats or a small measure of modest good humor fail to arrive after all this humiliation. This is especially perplexing given that Singularitarians seem to want to pass so desperately as serious scientists -- when scientists are, after all, among the most scrupulously caveating folks I know.

Beyond all this, as I have been reiterating repeatedly here for the last few days, the concepts to which Singularitarians make regular recourse in their discourse look deeply inadequate to me when they're not actually actively incoherent. I still await the sense that these folks take the particular embodiment of actually existing consciousness seriously, or register more of the actual diversity of capacities and performances that get described as "intelligence," or show more awareness of the histories of the value-discourses they appropriate when they start to go off on "friendliness," "unfriendliness," and the rest, not to mention demonstrating a little more awareness of widely understood and respected critiques of technological determinism, reductionsism, autonomism, unintended consequences, and so on. When they start barking about transcendental inevitabilities and number-crunching the Robot God Odds so solemnly I sometimes suspect I'll be left waiting for these little niceties forever.

About the skeptics, most of whom, we are assured, are sloppy-headed "literary types," Yudkowsky tosses his head with dismissive scorn:

"It's not that they think they have a theory of intelligence which lets them calculate a theoretical upper bound on the power of an optimization process. Rather, they associate strongly superhuman AI to the literary genre of apocalyptic literature."

It's simply breathtaking that Yudkowsky seems to think Singularity skeptics actually need to re-tool their understanding of intelligence to address the weird and wacky things Singularitarians claim in the name of their scarcely digested notions. The whole point is that if your discourse proposes that "optimization" spits out a Robot Overlord you are not making a claim that is exclusively or even primarily located under the heading of "computer science" -- it is doing work for you that is more like the work of apocalyptic literature. Network security discourse can cope with recursivity without becoming a cult or keeping a charismatic (?) guru in gloves and fans. It is the investment of projected (in more than one sense of that word) malware with entitative motives that takes us into the realm of quasi-religiosity, of collective dread and wish-fulfillment, stealthed as scientific objectivity. This latter investment renders Singularitarian discourse vulnerable to Superlative Critiques which recognize the cultural iconography that is getting mobilized to whatever ends and such critiques expose this mobilization. The confusion of warranted consensus scientific description with cultural mythology is not ours when we discern it, but Superlative as its partisans depend on it.

Of course, in light of all this, the endless self-righteous attribution of ignorant fashionable nonsensicality to anybody who refuses to give in to the reductionist triumphalist bulldozer of scientism (which is not, remember, consensus science itself, but always an opportunistic appropriation of science for parochial political and cultural ends) of the Superlative Technocentrics is just adding insult to idiocy.

The fantasy that the silly humanities people can be browbeaten into silence because self-appointed high priests of science (by which, flabbergastingly-enough, Yudkowsky seems to mean people like himself) can scrawl their equations on a chalkboard all the while oblivious to the metaphors that do so much of the actual heavy lifting in their arguments, or the fantasy that self-appointed technocratic elites with weird skewed priorities can deny the diverse demands of the actual stakeholders to ongoing and emerging technoscientific change all for "their own good" are the real frauds that demand exposure at this time in my view. Sensible and informed "literary types" have a role to play in the exposure of these deranging technodevelopmental discourses as well as sensible and informed "science types."