| |
| |
| |
|
Page: 1 2 3
Comments:
<0> <DanB@ef> werd
<1> The code at aima is garbage.
<0> <DanB@ef> aima?
<2> why? are you looking at the python code?
<1> rahul: Yes, the Python code I mean. I can't attest to the Lisp code.
<2> no surprise
<2> it's python code after all :P
<2> especially if norvig wrote it
<1> It's uncomplete Python code. I don't know why they even bothered posting it. It's mostly "cl*** AmazingAIThing: #todo"
<2> he considers python to be comparable to lisp
<2> haha
<2> nice
<0> <DanB@ef> norvig and russel?
<1> When it comes to clarity and ease of design, Python kicks the crap out of Lisp :)
<2> there's no need for one
<2> you made a ridiculous statement
<2> Lisp code is as clear as you want it to be
<1> That's only evidence of your inexperience or naivete.
<2> HAH
<1> I suggest you read http://norvig.com/python-lisp.html
<2> hahahaha
<2> that page, haha
<2> it's blatantly false
<2> norvig doesn't know python
<1> I'll give you credit, you sure know how to type "hahaha" a lot better than argue. When you get a job, I sure hope you have a better strategy when talking to your boss.
<2> it's quite clear from what he says
<2> "when"?
<2> Cerin: if you think you know python, you'll see plenty of factual errors on that page
<2> and, he glosses over basically all of lisp when he compares it to python
<1> The article's a little dated, but I don't see any glaring errors. It's a short comparison between the two languages. He's not going to post a whole book detailing the minute differences.
<3> lol @ "I sure hope you have a better strategy when talking to your boss"
<2> xeio: I tend to choose bosses that don't blatantly lie :)
<2> Cerin: there's factual errors about python all over the place, and he missed the whole point of using lisp
<2> which is that you can come up with better syntax for describing the program
<1> errors such as?
<2> when he talks about lists
<2> they're not comparable between python and lisp
<3> i find python to be "unfinished"
<2> oh wow, he updated the page
<3> heh
<1> rahul: uh, yeah, and he mentions that in the "gotchas" section.
<2> I guess enough people complained
<2> GUI libs are standard in python?
<1> "[Python] Lists are not Conses"
<2> is that why there are so many of them?
<1> Yes, Tkinter is included as part of Python's stdlib.
<2> haha Tk
<1> hey, he didn't say Python had a "good" standard GUI :)
<2> so we can ignore that difference
<2> and the whole make-hash-table thing is bogus
<2> it's trivial to create a hash-table literal initializer
<1> But you still have to make it yourself in Lisp. Python has standard syntax for that a hash/dict.
<2> "make it yourself"
<2> that's something you could do as an intro to lisp
<2> it's the first hello world
<2> the second being a custom html-generation syntax
<0> <Autre@ef> is this a bible thingy?
<1> Autre: Sorta. We're comparing programming languages.
<2> and I don't know why he says that m-v-setq is a bad thing
<2> especally since it doesn't use heap space
<2> and I guess having data types in the core language is bad
<2> but having nasty GUIs in the core language is good.
<1> "core language"?
<1> Tk is a package in the standard library. It's not part of the syntax or anything.
<2> so GUIs are more useful than data structures, according to norvig
<1> I don't see where he says that.
<2> you better come bundled with a GUI, but if you come with multi-dimensional arrays, that's bad
<0> <DanB@ef> how can a GUI be 'more' useful then a data structure, aren't they different things?
<1> Python has quite a few data types.
<2> norvig says that a built-in GUI is good, but built-in data structures are bad.
<2> even if that GUI is nothing anyone will want to use :P
<0> <DanB@ef> is this the same norvig that wrote a modern approach?
<2> yeah
<0> <DanB@ef> have you read that book?
<2> this is his self-justification for joining Google
<2> yes
<2> and PAIP
<0> <DanB@ef> he's joined google, and you read the whole book?
<2> yes
<1> DanB: rahul can't stand anyone saying Python is even remotely more useful than Lisp in anything, so he's spinning Norvig's words
<0> <DanB@ef> ahh
<0> <DanB@ef> hehe
<2> where am I spinning it?
<2> it says this exactly on his page
<1> Sorry, I can't find "gui is better than data types" anywhere in the article
<2> GUI not standard is bad
<2> data types in core language is bad
<2> this is #ai. get with the program.
<2> we're supposed to understand logical inference I thought
<1> That's one con about Lisp. He also mentions several good things about Lisp, and several cons about Python, such as its lack of macros and slow speed.
<2> for some reason packages in lisp are hard to use, but they're easy to use in python
<2> Cerin: ok, and I don't have any gripe about that, except that the macros thing has been faked in python
<2> since the data structures you get in the macro are so different from the way the code looks
<2> I wonder if CLPython gives python code decent speed
<0> <rkz@ef> python will never be as close to the speed of lisp
<0> <rkz@ef> because of the whole "batteries included" philosophy
<1> Python's real weak point, with matrix multiplication and array access, can be byp***ed by using packages like Numpy.
<2> I guess the fact that everything requires a hashtable lookup hurts it
<2> in lisp, object member access is just an array access
<2> accessors get optimized into array references
<2> not sure how easily that can be done in python... maybe it could involve some sort of scanning the code for common member accesses and putting them in an array instead of the hashtbale
<2> anything where the member name is hard coded would have that member put in the array, maybe
<2> you'd still need an entry in the hashtable for dynamic accesses, but that could be a forwarding pointer kind of thing into the array
<2> his main legit gripe against lisp seems to be that adjustable arrays are not the default
<2> and there's no :initial-contents to make-hash-table
<2> can you tell a python array to pre-allocate a certain amount of space but have a smaller number of elements to start with?
<2> like lisp's fill-pointer?
<0> <DanB@ef> bbl
<2> does python allow reasoning about type hierarchies, like "is this type a subtype of this one"?
<2> and then allow that to return "maybe"?
<2> for example, (integer -10 10) "may be" a subtype of (unsigned-byte 8)
<2> I still don't see where norvig says your programs are clearer in python
<2> he only says that it looks more like the pseudocode in AIMA and it's better than Java
<4> will tera-scale computing per single chip create a breakthru in ai, a powerfull optimization process?
<5> master icez
<6> weee
<5> hows u mate?
<5> im just a lil drunk
<5> :D
<6> jeez
<6> i'm good
<3> lol
<6> i thought you guys only drank tea up there
<6> :)
<5> well :P
<5> im drinking it now ;)
<5> so does that count?
<6> lol:P
<5> :)
<5> so what u been up2?
<1> rahul: No, I don't think Python lists can have memory pre-allocated. Their size is allocated as needed. And yes, Python has the standard function issubcl*** that can determine if one type is a subtype of another. Note in Python there's no real difference between types and cl***es.
<2> oh yeah, forgot that python's type system doesn't even have such concepts
<2> it's just a tree of types.
<1> rahul: Do you work with Lisp for a living?
<2> yes
<5> doing what rahul ?
<2> airline check-in management
<1> rahul: What dialect do you use?
<2> checking their reservation, printing boarding p***es, bag tags, etc
<2> Cerin: CL
<2> duh :P
<1> Well, there's Scheme, and a few others. I've only done a little Scheme myself, so I don't really know what Lisp programmers use.
<2> hah
<2> scheme just wouldn't be practical for this
<2> compilers aren't mature enough and it's not extensible enough without doing lots of extra work
<1> Have you ever looked at any of the cognitive architectures like ACT-R and Soar? I believe they're mostly implemented in Lisp.
<2> haven't looked at them
<2> heard of them
<2> I don't actively do much AI
Return to
#ai
or
Go to some related
logs:
#css
bofh-bash
#perl
freenode SET SECURE
https web.py
fedora make new partition
how to write a plugin for nagios
#debian
mencoder + sampling rate 8000 is not allowed in mp2
zmd snsl
|
|