Many of the target problems identified here (NLP, constraint solving)for which logic/prolog were used in the past have simply been replaced by brute force sat solving and deep nets: thinking as pattern matching.

I hear less and less about prolog used anywhere.

Edit: I'd like to hear why this is downvoted. Not objecting to the downvote, but would like a reason.

Here's some anecdotal evidence: google trends for "prolog" looks like this:

https://trends.google.com/trends/explore?date=all&q=prolog

Prolog has seen significant use at “3AM At My House, Inc.”

You’re right, we have a lot of new compute power that came online, intractable brute-force became the smart way. For now.

Most of my life we cheekily made fun of “mainframe computers the size of a room,” and here we are running GPU warehouses the size of the Vatican.

Datalog is quite hot still.

The associated book by the same name and author is very, very good. It's an excellent introduction to both the content specified by the title and Prolog. It shows the power of declarative programming in Prolog, where you basically right down the constraints of the problem you're trying to solve, and then it solves it for you.

In theory. In practice writing a bunch of constraints without understanding how prolog search works will get you terrible performance. It's essential to give constraints so as to match the search algo. This is straightforwad and not a problem at all, but it's not a step you can omit when faced with a large (large = anything non-trivial) search space.

That's true. Although, the book does address this in instructive ways.

Thank you for the great comment. I will go get the book now.

No problem! It's a great introduction to Prolog.

With recent progress in AI, this is surprisingly outdated. Several assertions are rather weak and not all that useful.

Outdated in what sense or how?

Thinking as Computation was developed from a course meant to introduce first-year students, from a wide variety of disciplines, to the topic. The prerequisites are thus very light and require little more than high school mathematics and no programming experience.

The author directly describes the scope and context of the book in the preface, and it was primarily designed to give students hands-on experience on the topics. Prolog turned put to be a successful choice in teaching students with no programming background and helped guide the course material.

The book isn't really an introduction to AI, although the topic lies within the larger umbrella of AI.

The Chinese room never made sense to me because, even though the person passing the arcane symbols through the door doesn't _understand_ what is being communicated doesn't mean that's not how it works in _our_ brain. Our individual neurons (or aggregate neurons depending on how they're working and firing based on symbol triggering) don't know what symbols they're processing either. So unless Searle is saying humans don't know any languages on some fundamental level, it's kind of useless because then we'd just end up admitting (something unsavory) that nobody knows anything barring some homunculus in our brains or even worse, we can never truly have free will. Those are scary implications.

To me, having part of the chinese-speaking-ability be in an almighty dictionary always looked like a sleight of hand; surely if you’re talking to the whole system that is the room, the person, and the dictionary, it is the system as a whole that has intelligence and knows how to speak chinese? It takes advantage of our identification with just the person to distract from that.

Wow. I never thought of it from that light. Honestly, that is a problem. You're smuggling in pre-formed language into the thought experiment with its own rich meaning.

Haven't read the top, but increasingly I'm obsessed with something that sounds similar: "Life as a computation."

All the adaptive properties of a living organism can (very rudely) be imagined as "if" statements. To us humans[^1], a "more complex" form of life is invariably one that can leverage more computation to ensure (reproductive) success. For example, a cabbage plant is more resilient to mild environmental and biological threats than a protozoa, because it contains "programs" for forming stems, root systems, energy-gathering leafs, bark and so on, not to mention their sophisticated immune system "subsidized" by the bigger amount of energy the plant can collect.

Then there is thinking: a tool that has allowed Homo Sapiens to work around their biological limitations and achieve a measure of comfort.

(Some) governments, religions, and other bits and pieces of culture present forms of collective thinking (computation). Humans are formidable as part of these structures.

Computers allow Homo Sapiens to move some of their life-sustaining computations out of their biology, which can be used for great evil[2], but are mostly used for the benefit of our species.

The practical corollary of this ramble is that we will likely demand much more computing power in the future.

----

[^1]: So caveat emptor, it's on the eyes of the beholder.

[^2]: Think, nuclear bombs or bioweapons.

These are some great slides. Thanks for linking this. Is there a newer version?