I came across with an interesting video today that displays experimental use of a bunch of tools this organisation has developed, which aid students and researchers with AI studies and experiments. The website is aispace.org, check out the video below:
Language is beautiful. It is one of the easiest methods of communicating something lengthy. As one may know, images express a thousand words, film may even express more, but both are ineffective when it comes to communicating something fast and immediately to someone (unless you have a polaroid of course).
English, particularly, is much more complex than I thought. As you may already know, I am rather _into _AI. One of the fields of AI is Natural Language Processing which involves speech recognition, syntactic and semantic organisation, separating words into parts of speech, language understanding, and finally speech generation, amongst others.
I genuinely used to believe there were only 8 parts of speech in English: Verbs, Adverbs, Nouns, Pronouns, Adjectives, Preopsitions, Conjunctions and Interjections. I stand corrected by Wikipedia. There are about 17 different types if we classify them correctly.
The open word classes:
According to Wikipedia, “an open class (or open word class) is a word class that accepts the addition of new items, through such processes as compounding, derivation, coining, borrowing, etc.”
The closed word classes:
According to Wikipedia, “a closed class (or closed word class) is a word class to which no new items can normally be added, and that usually contains a relatively small number of items.”
Now, I was not aware of such classification, but this for sure must have an impact on language parsers. I am currently starting to write my first Language Parser, and thought I would start out with an SQL dictionary structure. I will be posting my findings in this blog as I get along with this.
Today I read an article about A.I. on NewScientist called “Why AI is a Dangerous Dream”. I thought the article was partially biased. In my perspective the interviewee, a robotics expert named Noel Sharkey has lost faith in AI which can be confirmed by a statement on the article that reads:
I have dreamed of AI since childhood, creating flooders, scrollers, and chatter bots. I once developed a complete TCP/IP application using the MSNP8 to log on to the MSN Messenger network with a bot that would simulate intelligent talk. I have always been fond of the Turing Test. I even held a small conversation over email once with John McCarthy. I know this does not make me an expert in AI, however, it does make me an AI enthusiast.
There were two segments of the article that really put me off:
Anything that can be defined as a physical or logical entity or construct can be emulated. Anything that can be emulated could work equally or even be superior to the original. If a machine can emulate the logical processes of the human brain, then it can be said that the machine is intelligent.
AI is beautiful. It let’s us, as humans, test ourselves to our limit. It allows us to analyze how the human system works, and attempt to imitate our inner construction, our mind. It allows us to try to break the barriers, and build machines that are capable of so much more than us, and so much faster.
I am aware that AI often becomes somewhat of a cult. Sometimes it attracts the same type of people that follow all of Steve Job’s life events, spam online forums and blogs that post articles against the iPhone, and get hard-ons at a keynote. I am aware that AI can be a “believers” dream, but I am also aware that not everyone is like this. Not everyone takes their beliefs to a “religious” level. I am aware that to accomplish goals you have to be down to earth.
For starters there is Strong AI and Weak AI. When you talk about AI, you should generally make a point as to which type of AI you are talking about. AI related to specific tasks, or AI related to the reproduction of general human intelligence. It makes a huge difference to an article about the topic.
I believe in both. The proper coordination of the different Weak AI segments _can _lead to a fully sentient being. And to study one topic, you must be knowledgeable in all surrounding topics.
If the human mind is nothing more than neurons passing on electric signals which can be described as thinking, and thinking is the mechanism that allows us to communicate, and everything we perform in the world is a form of communication, then why shouldn’t a computer be able to be “intelligent”? A computer after all has a heart (PSU), a brain (CPU & HD), a face (LCD), ears (Mic), a mouth (Speakers), eyes (Camera), and can effectively move and communicate through different mechanisms such as robotic limbs, wheels, and other accesories.
Human beings consistently change their thoughts, ideas, knowledge and personality in the same way a computer program could re-compile itself to meet new standards and “personality” as has been described by Matt Knox in this article about his days as an adware author.
So we have also got that point covered. We can interact hundreds of miles away at an office using a coat rack on wheels as described in this article.
I am aware that these are not true applications of AI as a whole. But we have strived in different areas, creating artificial limbs that move like an animal’s paw or a human’s leg or arm. We have created programs that can re-compile themselves to allow different circumstances to occur, or to “evolve” as we would say. We have created robots that can detect surfaces and objects and go around them. We have stuck computers on coat racks with webcams, microphones and speakers to be able to “live” in an office from hundreds of miles away. We have created chatter bots that are close to beating the turing tests. We have achieved a lot in speech understanding and generation in the past years.
This leads me to say that even though we have not achieved a real strong AI system, we are certainly on the path to producing great results, and who knows, maybe in a few years we will have discovered the pathway that will lead us to develop a remarkable electronic clone of us.
I believe alchemists have ever pursued the dream of living forever. But A.I. has given us the dream of creating intelligent copies of us. I believe that this dream will eventually allow us to download our minds into intelligent beings, technically allowing us to live forever. And while we might have not found the philosophers stone, we will have found the elixir of life.
My sister stumbled upon a very interesting website today, and so I would like to share this with my readers (and listeners).
This is a website related to music. In it, you have many videos available which produce sounds, and together, or one by one, create stunning music, in B flat.
I have to say, my utmost respect for those who created this, it is a brilliant experiment, and just stunning that you can play around with music all in one same scale in this way.
Check out the website right here: http://www.inbflat.net/
A few years ago I dreamt of being able to have all this connectivity, lets say, things being published here, automatically-published there, and elsewhere. I knew we were getting there, as I have been configuring my services for the past few months, but I didn’t know we are so advanced.
If I made a rundown or a graph explaining the connectivity between social media sites I currently have, it would grow too large, or the connections would not exactly be easy to draw. * Update: Well I’ve decided to draw it anyway:
For instance: My Youtube is posting to facebook, google reader, twitter and two of my blogs. One of those blogs is lemiffe.com which when receiving a post will alert ping.fm which will then release a post of my new blog post to facebook, twitter, my blog at blogger, myspace, google talk’s status, and quite a few other sites.
Now, the only problem I currently see is the level of redundancy. Sites talked to other sites, which in turn talk to other sites, so whenever these sites provide much more connectivity to other services we might end up in a cyclic-post system in which one service posts to the other which in turn posts back to the first one initiating a never-ending cycle in which the user would have to stop one or the other service manually.
I really believe more insight should be applied, like, verifying the other sites connectivity and checking it will not repost to a site the first site has already posted to. This can be done via pure XML transactions between both sites, I mean, thats what web services are for, right?
Another thought I came up with is a Centralised User Account Service, which deals with the connectivity between users of different sites. Lets say: Gravatar meets facebook and ping.fm. All the services first verify the users information on the Centralised service, then they verify which permissions the user has applied concerning posting, which pages must the service send the new update, video or blog message to, and negotiate it only with the Centralised site, which will in turn send off the appropriate XML for all other “connected” sites.
It’s just a thought, Security-wise I am not sure where this would lead to as very tight security would have to be placed. Gaining access to the account of a user at the Centralised Site would be, well, Armageddon for that user. So probably a tighter security would have to be placed, with 128 big encription, SSL sockets, MD-6, who knows what else, but it’s just a thought anyway.