Cheeky - It's Australian for Artificial Intelligence | Hacker News Reader

Cheeky - It's Australian for Artificial Intelligence

theglobalmail.org

65 pointsColinWright31 comments

Salon published in 2003 a great series of articles by John Sundman about the Loebner Prize. He describes a two-time winner, "ALICE," in a way that doesn't inspire awe for the creator's programming achievements:

"""Wallace’s theory of A.I. is no theory at all. It’s not that he doesn’t believe in artificial intelligence, per se; rather, he doesn’t much believe in intelligence, period. In a way that oddly befits a contest sponsored by a bunch of Skinnerians, Wallace’s ALICE program is based strictly on a stimulus-response model. You type something in, if the program recognizes what you typed, it picks a clever, appropriate, “canned” answer. ... There is no representation of knowledge, no common-sense reasoning, no inference engine to mimic human thought. Just a very long list of canned answers, from which it picks the best option. Basically, it’s Eliza on steroids. ... And this strategy works, Wallace says, because that’s what people are: mindless robots who don’t listen to each other but merely regurgitate canned answers."""

http://www.salon.com/2003/02/26/loebner_part_one/

If Wallace's theory is correct, then there is no reason for me to consider it, since it is just a canned response devoid of thinking, and my response would be similar. If his theory is false, then there is no reason for me to believe it.

The only logical response to his theory is to ignore it.

If Wallace's theory is correct, your internal monologue about believing or ignoring his theory is an illusion; the only logical response to his theory is to trot it out in conversation if it seems clever, and otherwise ignore it.

Tell me more about mindless robots who don’t listen to each other but merely regurgitate canned answers?

> ...mindless robots who don’t listen to each other but merely regurgitate canned answers

This only serves to sweep the definition of intelligence under the carpet. What is the algorithm by which people determine the most appropriate canned answer? The Chinese Room argument is similarly flawed.

I am by no means as smart as these people but I did recently made this observation myself while visiting Google+. I try to avoid political discussions but someone always finds a way to insert them. This time, as I glanced over one such discussion it made me realize... People don't think, they just repeat things that have been told to them, fact or not, as long as it sounds like a good fit. No one involved in these discussions is going to convince the other to change their mind. AIs should be this easy!

That is very true. Most of what I say is phrasing or idioms I've unconciously absorbed. However I obviously don't randomly repeat anything I hear. The things that I absorb are ideas that resonate with my existing world view, which has been crafted by original thought over a much longer time. So behind those memes and tired phrasing in the arguments you heard there is porbably a kernel of unique thought.

This is why I prefer online forums and other forms of asynchronous communication over face to face debates. The former allows me to form an opinion - which can often change as I try to articulate myself. The later relies on my initial reaction, which is always much less refined.

I think it's a bit unfair to say he cheated. He didn't break any rules, even if his program wasn't really in the previous spirit of the competition - but it also sounds like they needed a bit of shaking up anyway.

From the Loebner Prize page: "The winner of the annual contest is the best entry relative to other entries that year, irrespective of how good it is in an absolute sense". Sounds like his entry certainly was deserving of the annual contest, even if it's not a general AI solution.

Personally, I would agree with you.

But Hutchens' own paper was called "How to Win the Turing Prize By Cheating" -- http://www.csee.umbc.edu/471/papers/hutchens.pdf -- so I guess reasonable minds differ.

It's just his light-hearted way of calling out the Turing test, I think.

To me the Loebner prize does a lot more to teach us about humans than about how to create AI. The success of ALICE showed us how much of our conversation is just based on regurgitating canned answers based on queues, just look at how drunk people converse. HeX showed us how much we think we communicate when we're just exchanging banter.

This horizontal scrolling thing is a real bomb.

I'm a big fan of horizontal scrolling as it is the only scrolling my Apple Mighty Mouse will do these days!

It's so bombular that it doesn't even work on the outdated version of Chromium I'm running at work.

Isn't it? I was briefly using all of the space on my widescreen display while browsing. Certainly an improvement over vertically-scrolled websites where margins are often just blank space.

It looks like you still get vertical scrolling on mobile.

Why oh why do you browse with your browser maximized?

I could have viewed this page maximized and just about read the whole thing without scrolling at all, but how would that be preferable?

You might fit more text on the screen but you don't read the whole thing in parallel, you slowly read paragraph after paragraph and what you gain in not having to scroll you loose in having to move your eyes and head instead, running maximized you just waste screen real estate.

I would have expected for this implementation to break down with different browser sizes but it reflowed well and seems to be well thought out. But I'm very skeptical towards such a radical change unless there is a real benefit to it. The benefit in this case is that you get reasonably sized columns (that are easy to read), which is kind of hard to achieve with vertical scrolling (I haven't seen a single implementation that works nearly as well as this one).

So I applaud their efforts, but there should be a manual fallback (couldn't find one), because I'm confident that it doesn't work as intended everywhere.

"Why oh why do you browse with your browser maximized?"

Because white space is great.

There's a problem with people who need bigger fonts - on chrome on a macbook the text falls off the bottom of the screen and there's no way to read it. And frustratingly there's a bunch of space up top (Apple menu bar; chrome's weirdly large titlebar; chrome address bar and bookmark bar; then there's a white stripe; and then there's the Globe and mail banner. At the bottom of the screen there's another grey bar (Our audience is our only agenda).

Using Chrome's frustrating ctrl+ feature ("zoom everything", instead of "enlarge text") means that the user interface stuff takes up even more space, and there's less space for text.

People may know me as being vehemently against this type of website - "just give me the text with nice css" - but I actually really liked the experience, even with these problems. (I also liked the photo essays.)

You can get more vertical space by going fullscreen (I had to on my netbook, after zooming out a bit).

Meh, I found it difficult to read. I wanted to skim past the images and intro fluff...but had a hard time finding the real content to read.

For people with poor vision like myself it breaks even further. I can't ^+ to make it larger without moving words off the screen and I can't use my readability JS script because their HTML is borked.

Bomb as in "it's da bomb", or as in "it bombed"?

Or as in "bombe"? http://en.wikipedia.org/wiki/Bombe

On a retina iPad it was the nicest horizontal scroll page I've seen. I'd be perfectly happy to read any text like this.

I think they only tested it with Macs (where it is a joy to use). That’s just plain stupid.

The article does use the word "cheat" and the winner himself said he cheated and hacked, but actually he didn't cheat at all.

Cheating would be using a human to generate the responses and pretend it is a computer.

Reading the page with zoom is impossible, the page is inaccessible to me.

Jason Hutchens wrote a paper about his winning entry (linked in the OP), which is just as interesting as the OP. The relevant discussion starts from Chapter 7.

http://www.nyu.edu/gsas/dept/philo/courses/mindsandmachines/...

I wouldn't be able to read it at all had I not used View/ Style/ User mode in Opera, turning off all the "smartness" of the formatting.

Chris McKinstry was also fed up with the Loebner Prize, but instead of making a chat bot to break it, he invented a less subjective intelligence test, the Minimum Intelligent Signal Test (MIST) https://en.wikipedia.org/wiki/Minimum_Intelligent_Signal_Tes...

It accords with my (disappointing) impression from my undergrad AI class, that while these systems are certainly "artificial", the "intelligence" tends to lie completely with their designers.

Reply on news.ycombinator.com