Comments on: Ethics for machines

By: Indi in the Wired » … boldly to go where no being of a race of Earthly heritage has gone before!

Tue, 23 Feb 2010 00:13:46 +0000

[...] bringing this up is because AI blogger J. Storrs Hall brought up the topic in a recent post about rule-based AI. Hall actually went a step further, beyond the split infinitive, and looked at the second part of [...]

By: J. Storrs Hall

J. Storrs Hall — Fri, 19 Feb 2010 16:20:35 +0000

BBBeard: Smirk. Originally I had written “probably-better-than-state-of-the-art” as a single hyphenated word. That’s because I had a specific level of linguistic competence in mind, which I was sure was adequate for the reasoning I was talking about, but which I was not sure was captured by the state of the art. Reattaching “probably” leaves a sentence which has a similar, but not the same, meaning.

By: John Donigan

John Donigan — Thu, 18 Feb 2010 19:21:03 +0000

Actually,”Doc” Smith would have written “to go boldly.”

By: bbbeard

bbbeard — Thu, 18 Feb 2010 18:52:43 +0000

In re “cognitively expensive signaling behavior”: I’ll agree we’ve lost the war on split infinitives. We’ve also apparently lost the war on cursing in public, but that’s another matter….

The problem is that this leaves me, the beneficiary of an above-average education in the 60′s and 70′s, cognitively impaired. I find the split infinitive as jarring as mixing up “I” and “me”, or mixing a plural subject and a singular verb. Which is to say, yes, the meaning is unambiguous, but unambiguity of meaning is not the only purpose of grammar. Standardized grammar, like standardized spelling, increases comprehension and the rate at which we can assimilate information. Throwing in misspellings and solecisms may help a writer of fiction establish a style or build a character, but for most writing it is inappropriate.

Poor grammar reflects disorganized thought. Consider the sentence, “You can’t even reason about whether the Star Trek example is right or wrong without understanding language at a probably better than state of the art level.” What does “probably” modify? Apparently the author believes it should modify “better”. But a moment’s reflection suggests that the expression of doubt would be more effective if “probably” were used to modify the verb, i.e., “You probably can’t even reason about whether the Star Trek example is right or wrong without understanding language at a better than state of the art level.” This is not the only thing wrong with this sentence.

But such is the blog-oriented culture in which we live… or, as you young people say, “…which we live in”.

BBB

By: Mikee

Mikee — Thu, 18 Feb 2010 17:46:07 +0000

“…[T]o boldly go where no man has gone before” could have been changed to indicate earthlings, humans, Terrans, or “meatsacks” rather than “no one.” I’d have gone with “no humans” or “where humanity has never been before.”

Then again, I watched the show for the women in tight, short dresses. And green skin.

By: Instapundit » Blog Archive » J. STORRS HALL: Ethics For Machines.

Instapundit » Blog Archive » J. STORRS HALL: Ethics For Machines. — Thu, 18 Feb 2010 17:36:15 +0000

[...] J. STORRS HALL: Ethics For Machines. [...]

By: Stephen Reed

Stephen Reed — Thu, 18 Feb 2010 03:12:37 +0000

Hi Josh,
Excellent post on some issues in natural language processing that are dear to my heart. From the view point of Construction Grammar theory, I suppose that there is a semantic difference between the phrases “to boldly go” and “boldly to go”. The latter latin form, in my opinion, slightly reduces the “boldly” emphasis by moving it one word distant from the word “go”.

As you indicated in the post, simple grammar rules, e.g. phrase structure rules, are not sufficient to capture the compositional and idiomatic semantics of natural language. Construction Grammar directly addresses that issue. However, Construction Grammar merely transforms the problem domain – in that there exists no concise rule set to describe English, rather there are a great multitude of linguistic constructions that English speakers have inductively generalized from spoken and written instances. I’m working on figuring out of few of the most applicable English constructions for dialog.
-Steve