Future of Coding • Episode 76 Computing Machinery and Intelligence by Alan Turing With special guest: Felienne Hermans Hey, everyone? We've had this community for, what, like, 8 years now? We've shared a lot of great links, had lots of discussions. Surely most all of us know the name Turing, know about the Turing test. Heck, those Hollywood tastemakers put "Eggs" Benedict Cumberbatch in a movie called The Imitation Game, and it did numbers! Turing is top-tier pop culture for our field. So… why the hell doesn't anyone ever say, "This paper is proper messed. It's very very very very bad." Because, now that I've read this paper, I'm cursed! If anyone mentions Turing near me, I won't be able to resist the screaming. This lil Alan of all time has taken on an entirely new texture in my life. And if you don't know what you're in for, well, I cannot wait to welcome you into this new baffled, corrupted awareness. Now, I should say it was truly an honour to have Felienne Hermans, author of one of our all-time favourite papers, A Case for Feminism In Programming Language Design, join us. This episode format — a guest who discusses a work with us, but not their own work — is something Jimmy has been encouraging us to try for a while now, and I think it turned out fabulously. And we couldn't have asked for a better first guest — or a worse first work. Enjoy!

I have heard it said that when people are impressed by a pig dancing, they are not impressed by how well it dances. They are impressed that it is dancing at all. Similarly, I have not heard this paper praised for its depth, rigor, or comprehensiveness. It is usually praised for the fact that he was able to think about these ideas at all given the extreme limitations of the computers that were functioning at the time. The discussion of rules, and the fact that there must be rules for interpreting them, reminds me of the excellent discussion in Formalization in Program Development (1982) by Peter Naur. Peter Naur was also skeptical of Turing's thinking on AI. See Thinking and Turing's Test (1986).

yes but can the pig be of either sex?

If you just can't get enough of the FoC podcast (whatever the hell it is), there's a just-released episode you can lis- No, no, not the Turing one, silly bean, that came out like 16 hours ago. Old news. There's a just-released bonus episode where Jimmy and Ivan (whoever the hell they are) talk about… the Turing episode. And Blue Prince. And end with a C02 joke. How's that for theme and variation. Give us $5, enjoy.

Lu disappeared, though!?

Yeah, Lu turned into a ghost halfway through the Turing ep, which is why their closing thoughts appear at the very beginning.

“I am like an elephant with a capable brain. I have a soul.” — Might be an all time best out of context quote.

Love love love this ep 😄 It was savage, but necessary, to create space for all the voices alluded to in the conclusions. Look forward to hearing more! I’ve read “Computing as Theatre” a while back, and I really liked how it pulled me out of my “Technical” mindset to look at what a computer was. Another time that happened was when I went to a meetup of uni students who were studying Arts, and as ppl who don’t program, they were writing poetry using code - the way they were using code was quite refreshing to see, unshackled by a compiler/interpreter. My Technical brain kept asking what’s the point of that? But my curious inner child was tickled by the experience 😄

On the bonus episode you discussed what other papers to read. I wonder whether The Bitter Lesson might evoke strong and interesting reactions. While nominally AI-focused, it has ended up being more broadly influential recently. I've been arguing with it in my head for a while. Might be too short to generate one of the sprawling 2+hr episodes I enjoy, though.

Listening to this episode as someone who had heard of but never read that paper, I found it both interesting and fun, but also disappointing. Reviewing a 75-year-old paper should be an exercise in history of science if you want to learn something from it. Example: you refer to "peer review" a few times, but don't seem to be aware of its history. Peer review was invented in the 1950s. Unless "Mind" was a pioneer in this development, Turing's paper was most probably not peer reviewed. Which would easily explain many of its shortcomings, because before peer review, authors didn't get critical feedback before publication. The standard procedure was (1) author submits manuscript, (2) editor accepts or rejects, (3) a technical editor revises the text, checking for typos, bad grammar etc. but also proposing more fluent phrases, (4) author accepts or rejects each of the proposed changes, (5) paper goes to print. Perhaps it also matters that academia was much smaller back then compared to today. It is safe to assume that Turing, the editor of Mind, and everybody cited in the paper knew each other and met regularly in conferences etc. Most of the readers of the day would have been part of that club as well. Written papers were part of a multi-channel debate and were seen as a recording of the most outstanding and lasting contributions. Another interesting historical perspective would focus on Turing's unusual status as a war hero, an academic celebrity, and a criminal (for being gay). I am sure that historians have explored this perspective, maybe even taking into account this specific paper.

i speak for myself but i personally don't care what reviewing an old paper "should" be

@Konrad Hinsen "Most of the readers of the day would have been part of that club as well." This feels like a job for "Deep Research," very curious what citations it will come up with.

@Konrad Hinsen Very well said. I suspect the attempt to avoid hero worship is being taken too far and turning into gratuitous denigration.

There's clearly a values mismatch here. That's okay, so long as nobody insists that their way is inherently "correct" for one and all.

@Ivan Reese Yes, there's a values mismatch, but I'm not sure how much of that is based on philosophical differences and how much is based on different understandings and interpretations of the facts. I say with all due immodesty that I'm a reasonably bright person. I can understand most of his important ideas and many of their implications, but I cannot fathom the thinking or the genius that was necessary to create those ideas given the intellectual foundations that were available at the time. I believe it is a fact that Turing made significant, lasting, fundamental contributions to our understanding of the nature of computation. I don't worship him, but I have a deep respect for him, and I consider his arguments carefully even when I disagree with him.

Absolutely should not have dived down this rabbit hole... But I did find one tasty treat down here... A 2021 problem set featuring three provocative quotes and asking the question, "The perfect simulation of intelligence is intelligence. ... Is that true, though? In a short essay, explain why or why not." https://philosophy.tamucc.edu/courses/spring-2021/minds-and-machines/problem-sets/ps-02 Though "the perfect simulation" goes too far, we can daily confront "a fairy iffy and sometimes surprising simulation". But the best bits... Robin Gandy, 1996:

The 1950 paper was intended not so much as a penetrating contribution to philosophy but as propaganda. ... I can remember him reading aloud to me some of the passages--always with a smile, sometimes with a giggle. Some of the discussions of the paper I have read load it with more significance than it was intended to bear. I shall discuss it no further.

John Haugeland, 1997:

Why single out talking for special emphasis? ... Talking is not merely one intelligent ability among others, but also, and essentially, the ability to express intelligently a great many (maybe all) other intelligent abilities. And, without having those abilities in fact, at least to some degree, one cannot talk intelligently about them. That's why Turing's test is so compelling and powerful.

Daniel Dennett, 1998:

I am confident that no computer in the next twenty years is going to pass an unrestricted Turing test. They may well win the World Chess Championship or even a Nobel Prize in physics, but they won't pass the unrestricted Turing test. Nevertheless, it is not, I think, impossible in principle for a computer to pass the test, fair and square.

Curt Jaimungal had a long conversation with Dennett shortly before he passed away .

None of us said Turing didn't make significant, lasting, fundamental contributions. We just said his paper and the arguments in it are bad, judged by our own personal standards and tastes and from a modern perspective. I get that this bothers some of y'all in the "have respect for the greats of our field" crowd. This podcast isn't going to scratch your itches.

I am actually not very interested in judging Turing posthumously. For me, "Turing" is a name from history. He died before I was born. I have no emotional relation with people I never interacted with. What I am personally interested in is to untangle the contributions from Turing as an individual, the state of academia in 1950, and the interaction between different disciplines around the emergence of a new one. For example, the obvious mistakes in the paper illustrate the value of peer review, which I consider one of the most important inventions in academia (even though its implementation needs an update). Another example: the apparent lack of communication between mathematics and early CS on one hand, and philosophy on the other hand. As you point out, it seems weird not to have had Wittgenstein in the loop. Was this an accident? A matter of personal relations? Or isolationism in each discipline?

Or just not being in the same room at the same time — it's easy to forget that that was requisite for a conversation. 👻 Having been at least somewhat impressed by the paper as young person, revisiting it, well, "bloggish" captures the spirit a bit, but basic can be captivating when an idea is new.

> Talking is not merely one intelligent ability among others, but also, and essentially, the ability to express intelligently a great many (maybe all) other intelligent abilities. Thanks for sharing this quote. This is the thing I kept thinking about during the podcast--and I disagree pretty deeply, at least when it comes to the "maybe all" part. The ability to read a room or read a person; to know exactly when a gentle touch on an arm will settle; to know when a child needs boundaries versus empathy; to know how to silently de-escalate a fraught situation simply by how you move your body...these are all very much forms of intelligence. Call it "social intelligence" or "emotional intelligence" if you want; the scare adjectives are just that.

Man, my first thought about language (and a good indication that it's to late on this side of the earth) is to wonder whether how often when we ostensibly talk about things whether intelligently expressing things... I mean consider Gianluca Gimini's bicycles... Here... https://www.wired.com/2016/04/can-draw-bikes-memory-definitely-cant/?utm_source=chatgpt.com

@wtaysom Could you elaborate? I'm not sure what you mean.

Ever met someone who could fix just about anything, but wasn't much for talking? (Or who had politician/MBA disease: could talk their way out of anything but never seems to actually...do...much?)

Oh, I mean in a much as most people cannot clearly imagine even the basic topology of a bike, I think when we push language, we're mostly not particularly robust to expressing ideas, but we make due by virtue of having shared context or getting on the same wavelength as the person we're talking with. Something like that.

That makes sense. I see how that perspective could help to explain how often Eliza has passed the Turing test, sometimes accidentally. The human could imagine a shared context in order to make "sense" out of the dialogue even if there is none.

I tried to pull in some of the historical elements by looking at what contemporaries thought of the paper. W Mays response was the most interesting. I later found out it was actually written the same year as Turings paper presumably to appear in the same issue. Gilbert Ryle was the editor of Mind and asked Mays assuming he’d be sympathetic (according to Mays reflection 50 years later) https://www.jstor.org/stable/3747787 It’s definitely true that peer review was not a thing at Mind. Gilbert Ryle was the sole editor and had sole discretion.

He read all contributions himself and made all the decisions by himself, without seeking advice from anyone else, except, as he told me, in the case of papers on technical issues of logic. He also carried out all the business of proof-correcting and the like by himself. He also gave his reasons for his decisions. When I took over I did the same in spite of some advice from elsewhere to the contrary. While contributors do not always agree with the reasons offered for the rejection of papers, most appreciate the fact that they are given, and Ryle's practice in this respect was one that I gladly continued. (https://shs.cairn.info/revue-internationale-de-philosophie-2003-1-page-5?lang=en)

I considered saying as much in our conversation, but as I’m sure you all saw, I said quite a bit there. I have a tendency to dominate conversations, I’m afraid I did that a bit much in this episode (though I try to justify it to myself by seeing myself as a representative for the text, I really did try to give turings arguments). I do get the criticism that we not respectful enough. For my part I try to give the arguments as they actually appear in the paper. I try to take the paper as a stand alone work whenever possible. Turing is great. If we had instead read On Computable Numbers, with an Application to the Entscheidungsproblem, we may have just being talking about how clever the whole thing was. Assuming we took the time to understand that dense paper. But, generally, I don’t think papers like that will make for good quick discussions. A paper like that needs a lecture and whiteboard and all that. Some of the best papers out there are probably beyond us to do in this format. We could definitely try, but I’m not sure they’d turn out great. I don’t know that we could ever take the historical approach and read the source material. We certainly could have read the loads of secondary literature on turing (I’ve read a handful). There are some that deal with all the issues we talked about. Some that try to clean up his arguments, some that don’t. But trying to get everyone to read all of that and talk about it all is very hard. One thing I’d sincerely be interested in is for people to write (or record [happy to do that with people], but I’m assuming writing is easy for many here) responses to the episodes. I’d be happy to find a venue to publish them. Happy to respond to them. I see all of this as an endeavor to learn how to read these text, how to talk about them, how to criticize them. All that said, this paper was not a great paper. I couldn’t find any evidence that Ryle and Turing were buddies, but Ryle was certainly sympathetic to the behaviorist project of Turings having published the concept of mind a year earlier. It’s also interesting that he didn’t publish W Mays response. I truly recommend reading it. He gets so much wrong that Turing got right, and so much right that Turing got wrong.