[mike777]

Google users promised artificial
intelligence
The Guardian May 23, 2006
*************************
A search engine that knows exactly
what you are looking for, that can
understand the question you are
asking even better than you do, and
find exactly the right information
for you, instantly is a future
predicted by Google. Google
co-founder Larry Page said one thing
that he had learned since Google
launched eight years ago was that
technology...
http://www.kurzweilai.net/email/newsRedire...ID=5572&m=22848

[Sigi_BC84]

mike777, on May 29 2006, 07:18 PM, said:

A search engine that knows exactly
what you are looking for, that can
understand the question you are
asking even better than you do, and
find exactly the right information
for you, instantly is a future
predicted by Google.[...]
http://www.kurzweilai.net/email/newsRedire...ID=5572&m=22848

It's sad that Google is already at the point where they are spreading similar kinds of marketing hype as Microsoft.

For a start, the search engine would actually have to be capable of understanding the searched content to even get close to the position of "knowing exactly what [the user] is looking for". I'm not sure how that second step would be achieved (some new kind of mind-reading device, probably).

Sadly, when the founders of Google "predict" such developments it is taken for real a lot more easily than if some renowned researcher predicted the opposite (or admitted that it's really an open question yet if such can be done).

True, Google does a great job in presenting the user with what he most probably was looking for in the first place, but that is not an oracle they're employing there. Merely a really smart ranking algorithm with a lot of engineering effort by quite a few really smart people thrown in.

Don't buy into the hype.

--Sigi

[mike777]

Thanks for reply.

As for mind reading device I assume you read my Honda post where without wires a human had a robot to move its hand merely by thinking it, yes? This is a long way from knowing what a human really means when the human dont know....but perhaps by then the machine will care less and the human will have the machine as part of it or the machine with bio parts of it? Mr. Steve Austin

[Elianna]

I used to work for a company that was trying to make these kind of searches. It was called Zapper, and it basically had this whole method of looking at the context around the words searched for, both in the originating text, and in the searched at text.

So it was ideally used not by typing in words in a search engine, but by opening a document, say an email, highlighting something in someone's email that interests you, and then calling the search engine.

It was pretty good, but the company is defunct now, I believe.

[helene_t]

Currently the bottleneck is mainly the user's capabilities of expressing himself. Even if he knows what he's looking for he may not be able to explain it due to bad typing or language skills. On the other hand, possibly a lot could be gained by letting the serach engine build up knowledge about indivual users.

What Elianna describes sounds cool as well. Even if the user is unable to explain what subject he's interested in, he can indicate it by pointing out fragments of existing text, written by others.

One day we will all have internet-connected brain chips that automatically provide information that is benificial to the user.

[Sigi_BC84]

I'm writing my masters thesis in a research department concerned with information retrieval and machine learning.

Ideas such as those mentioned by Helene and Elianna are actively pursued by a lot of the research that's being done.

For example, one student has designed an algorithm that uses the click behaviour of the search engine's users to refine the search. Others try to use natural language processing.

--Sigi

[luke warm]

helene_t, on May 31 2006, 03:28 AM, said:

One day we will all have internet-connected brain chips that automatically provide information that is benificial to the user.

wow.... you can then 'google' a subject wherever you are, whatever you're doing... the possibilities are limitless

[hotShot]

The main problem using web search engines is that you have to many hits.
You don't need help, if you hit only one page.
Here "AI" methods can help. If you search "Bridge" the results could be classified into groups:
Bridge (Building/Engeneeering)
Bridge (IT-Hardware/Network)
Bridge (Cardgame)
...
If you enter more than one word, it is possible to select the categorie you wanted, by looking on the overlap.

[whereagles]

mike777, on May 29 2006, 05:18 PM, said:

A search engine (...) that can
understand the question you are
asking even better than you do

When microsoft word tries to be more 'clever' than me, the result is usually me yelling at the computer

[mike777]

"I'm writing my masters thesis in a research department concerned with information retrieval and machine learning."

As in my other posts I would like to see more on the issue of how we end up defining the word "machine". I wonder just as we can debate where the head ends and the neck begins...I wonder if we will debate where the human parts end and the machine parts begin? Another issue may be if a machine is made from living parts that fit the definition of alive, is the machine alive?

I hope in your thesis you define your terms and speculate whether these terms may radically change in definition in the next 25 years. The Turing test was always interesting but I never quite saw where the word machine was defined or the word human for that matter. How many machine parts can a human have before we call it Borg anyway?

[Sigi_BC84]

mike777, on May 31 2006, 02:23 PM, said:

The Turing test was always interesting but I never quite saw where the word machine was defined or the word human for that matter. How many machine parts can a human have before we call it Borg anyway?

Turing himself defined an abstract machine (called "Turing Machine", quite surprisingly) and it was proved that this machine is as powerful as any general purpose computer. This only means that the class of problems this machine can solve does not increase if you make the machine more sophisticated.

You are usually taught about Turing machines in an undergraduate computer science course. It is a very fundamental concept because it shows mathematically where the principal limits of a digital computers are and that you cannot extend those limits by making the computer faster or more efficient in any way.

This is also the reason why you don't define "machine" over and over again when doing research in computer science.

Now it is completely undecided if the human brain is a digital computer (ie. equivalent to a Turing Machine) or more than that. Some argue that it is indeed equivalent to a digital computer (which would mean that there must be some way of simulating the human brain exactly with a digital computer). Others disagree with that position.

Computer science concerns itself with solving problems on Turing Machines (or their real world counterparts, being TMs with limited memory).

If we find a way some day to reliably connect brains and machines and it turns out that you get more than a Turing Machine that way, you will have to reconsider a lot of things. This depends entirely on how the brain actually works, and we still know next to nothing about this.

A good and easy to read book is "The Emperor's New Mind" by Penrose.

--Sigi

[mike777]

Ya I just reread that Penrose book. It was excellent. Let me state this book is written for us non Math majors but it has enough math in it for PHD's I bet .

He does discuss the Turing machine in detail. I notice you used the word "digitial" computer. It will be interesting if other types of computers redefine the word machine.

I note Penrose strongly argued that the human brain could not be duplicated or replicated by a Turing machine. Others argue that AI will truly be an intelligence in every sense of that word well before 2050.

[helene_t]

whereagles, on May 31 2006, 01:03 PM, said:

mike777, on May 29 2006, 05:18 PM, said:

A search engine (...) that can
understand the question you are
asking even better than you do

When microsoft word tries to be more 'clever' than me, the result is usually me yelling at the computer

Good point.

But when I mis-spell a Google search, it caries out the search as I spelled it and adds a "Did you mean ...?" link. Word procesors should do the same: indicate the place of auto-improvement with a color and writing in the status bar that I can press some key to accept the auto-improvement.

Btw, Pandora and Amazon get a picture of your taste and use that to recomend music rsp. books that you might like. I don't think they claim to know your taste better than you do, they just have quick access to music rsp book libraries, indexed by taste parameters.

Btw, this is the second time that I have to advise you to get rid of Microsoft. When do you join us followers of the Linux gospel, Whereagles?

[whereagles]

helene_t, on Jun 2 2006, 08:33 AM, said:

Btw, this is the second time that I have to advise you to get rid of Microsoft. When do you join us followers of the Linux gospel, Whereagles?

I *am* using Linux

I use it at work. Unfortunately, it's not devoid of some "aargh factors", though most aren't so exasperating as those of windows.

The version I'm using, Fedora 3.something, is actually quite buggy. Especially when it comes to interacting with applications. But I can live with it.

[Sigi_BC84]

mike777, on Jun 1 2006, 06:10 PM, said:

I note Penrose strongly argued that the human brain could not be duplicated or replicated by a Turing machine. Others argue that AI will truly be an intelligence in every sense of that word well before 2050.

Those scientists who believe in brain = digital computer are proponents of the so called "strong AI". Penrose indeed is in the other group of people, who doubt that it is that easy to get strong AI. (FWIW I'm on Penrose's side of the argument).

In any case I'm doubting that even by 2050 we will have understood enough about the brain and conciousness to even begin simulating it in hardware. This being only a gut feeling of mine I might easily be proven wrong, actually I'm hoping so. Having machine intelligence sounds exciting.
We will then see the technological singularity as well (also exciting ;-).

Having said that, all what is being done in computer science (and practice) today with "AI" in its name is much closer to bread-and-butter computer science than to any kind of "true intelligence" or brain simulation. Systems like neural networks which kind of mimic what we have found out about the brain so far are hideously hard to handle and for that reason being actually employed only in a few special areas so far. It's not that you can build one of these networks and "switch them on" and some kind of intelligence mysteriously develops. That is science fiction but has nothing to do with reality (as of yet).

I might also add in the possibility that brain and mind might not be the same thing. The brain could be merely a device enabling mind/consciousness, so by mimicking the (physical) brain you still would not achieve "mind" (this is somewhat along the lines of Penrose's arguments).

All of this is very fascinating in any case. I love speculating about it.

--Sigi

[Sigi_BC84]

whereagles, on Jun 2 2006, 11:59 AM, said:

I *am* using Linux

I use it at work. Unfortunately, it's not devoid of some "aargh factors", though most aren't so exasperating as those of windows.

The version I'm using, Fedora 3.something, is actually quite buggy. Especially when it comes to interacting with applications. But I can live with it.

A few remarks from a dedicated but not fanatic Linux user:

"Fedora 3" is the name/version of a Linux distribution (ie. install system, administrative tools and packaging of programs together with the Linux kernel), and not some "Linux version". This might sound like nitpicking but it's not.

Linux is the name of the kernel. The kernel is the thing most close to the hardware (it contains all the drivers that talk to the devices) and it also manages resources (hard disk space, memory). The version number of the kernel actually correlates to which hardware is supported.

The Linux kernel usually is packaged with the "GNU toolset" (hence some people prefer the name "GNU/Linux" over "Linux"), which provides for the actual Unix environment one is used to (the shell, many commands). Along come many other tools and applications such as the graphic interface and so on.

All of this is packaged into a manageable form by the Linux distributors (eg. Debian, Suse, Redhat, Ubuntu, you name it), each of whom follows their own way of doing things and preconfiguring the various applications.

The main problem you will be running into when using Linux is hardware compatibility, aka missing drivers. Vendors are beginning to jump the slowly rolling bandwagon and more and more provide Linux drivers, but it is still a pain in the neck running Linux on very recent hardware. Specially troublesome are any recent notebook computers.

Another sore spot is the lack of special applications such as BBO, which often don't get ported to Linux due to lack of market/demand. This is where the cat bites its tail, but luckily the "important" stuff such as web browsers, email applications, office suites is available in high quality (and mostly free!). So I'd say that for unassuming users who only rely on standard software Linux is fine.

I personally recommend Ubuntu Linux (which is derived from Debian Linux), which has a very dedicated and generous person backing it with a lot of money. They send you CD-ROMs for free, no strings attached whatsoever, if order them on their website. It installs nicely and works really well (I'm using it myself).

And yes, the "aargh factors" definitely are there. I'm quite knowledgeable concerning Linux and I've got my "aargh moments" as well from time to time. A big difference to Windows, however, is that there is almost always a way to find out exactly what went wrong and where it went wrong instead of having to stare blankly at a more or less meaningless error message with no log files, as it is usually the case with Windows.

Linux is a lot more secure. Basically you don't have to care about viruses and such. This goes a long way nowadays. Actually I might even say it could be the killer argument for many users in favor of Linux. In principle, Linux is vulnerable as well, and Windows in turn can be made a lot less vulnerable, but in practice, Linux fares much better in this arena.

So if you are one of those typical users without special needs, ie. you are happy with your word processor, email and web, I strongly suggest giving Linux a try. Then start bugging Fred about porting BBO ;-)).

--Sigi