classes at UFPE

[gusl]

Today I saw half of the first Information Theory lecture at the EE department. It was progressing at a good pace, but I had to leave halfway through because it conflicts with AI.

Then I went to the math department to see what classes they are offering. I'll probably attend one out of "Topics in Combinatorics", "Commutative Algebra", "Graphs and Algorithms". It may be too late to sign up properly, but I don't care too much.

AI was very entertaining again. Jacques surreptitiously gave us a survey of AI, from historical and methodological perspectives (including a nice cybernetics-like diagram of the concept of an agent), cognitive science, connectionist vs logical, only to ask, at the end of the class "but what IS intelligence?".
"Who is more intelligent: Ronaldo or Kasparov? A lawyer or a one-year-old? A doctor or a maid?"
Well, Kasparov was beaten by a computer in 1997, but soccer players as good as Ronaldo may not be built for another 50 years.
If the task is learning a language from scratch in a 10 year period, the one-year-old will do a better job than the lawyer.

Afterwards, I asked him for some professional advice, and he gave me what seemed like good feedback.
I also told him about my theoretical interests, and he said he has some kind of software project where a language needs to be developed, but his student doesn't have the kind of profile it takes to work out a formal semantics, so I may have a concrete project to work in.

Comments

[peamasii.livejournal.com]

The project sounds interesting. Have you thought about using 'lex/yacc', you can create your own program rules to teach the interpreter how to parse the input. I've envisioned using it for a token parser to recognize expressions in natural language and make flow decisions based on their order and recognition.

I don't know much about AI, but it seems writing a new language may be overkill if you're just trying to reach a specific goal (rather than an abstract framework).

[gustavolacerda.livejournal.com]

A language can be a specification language, a description language, a modeling language, etc.... and btw, this is probably not AI project.
But I actually have no idea what the project is about, so I can't judge his decision to create a language.

[adamas.livejournal.com]

Isn't dropping into classes fun? Sometimes I think it's a lot more useful than attending classes I'm enrolled in (or should say "should be enrolled in," all things told)... For one thing, I think learning comes a lot easier when it's natural and non-judged; seems most people tend to lose sight of this in favour of fretting over GPAs...

I have to agree with your above commenter; I think developing a whole new language seems a bit overdone. I wouldn't pick lex/yacc, but what's wrong with the typical AI languages (Lisp, Prolog, Scheme, et al.)?

FYI, I'm insanely jealous -- your AI class sounds amazing. :)

[gustavolacerda.livejournal.com]

see my comment above

[gustavolacerda.livejournal.com]

Isn't dropping into classes fun? Sometimes I think it's a lot more useful than attending classes I'm enrolled in (or should say "should be enrolled in," all things told)... For one thing, I think learning comes a lot easier when it's natural and non-judged; seems most people tend to lose sight of this in favour of fretting over GPAs...
Sure, sometimes. Especially if your goal is to be inspired. Unfortunately, I haven't shown the ability to commit to working in classes that are not for credit. I've probably been in the reward system for too long. But now it's my chance to commit for no GPA reward.