AI designs its own video game →
newscientist.com
Interesting article about AI system Angelina that is capable of generating video games on its own using evolutionary algorithms.
March 2012
7 posts
Google Knowledge Graph Could Change Search Forever →
mashable.com
Google now wants to transform words that appear on a page into entities that mean something and have related attributes. It’s what the human brain does naturally, but for computers, it’s known as Artificial Intelligence.
Although it sounds less fascinating than some kind of flying robot, this type of research can actually have a great influence on our productivity by improving the process of finding the information. Google is certainly not the first attempting to extract more knowledge from web pages than keywords but given the resources Google has, both material and intellectual, it is reasonable to expect practical results soon (1-3 years).
Bro , thanks for your blog , half of your blog will be on my presentation next week. I was wondering if you know any good examples or any advances of AI on medical fields ?
There is a lot of AI in medicine, mainly used for diagnosis and medical image processing. For example, diagnosis of pancreatic cancer (the one that killed Steve Jobs) from computed tomography imaging.
Among recent demos of AI in medical field the one in this video is perhaps the most spectacular.
http://worldsciencefestival.com/videos/triage_by_transistors
There are a lot of hopes for future application of Watson in medicine. Here is the video about it I posted before:
http://aiblog.tumblr.com/post/7838892598/watson-computer-comes-to-the-university-of
Good luck with your presentation.
Sentiment Analysis for Twitter →
tweetfeel.com
Want to know what people think about a particular issue, a product or a public person thеn sentiment analysis is the technology you might be interested in. The link above demonstrates how sentiment analysis is applied to Twitter feeds to mine the opinions about Obama.
The task of sentiment analysis is to determine the polarity of text, such as negative, positive or neutral. The basic way to do it is to look for words that express the corresponding attitude. Example of positive words are “good”, “satisfied”, “happy”; negatives are “fail”, “sad”, “broken” and so on. The trick is to collect a large vocabulary of such words automatically and to consider the context where they are used in.
Thank you so much for everything you post. I have a VERY VERY VERY VERY IMPORTANT presentation about AI and you blog really helped!!
Glad it helped. I haven’t been updating AI blog for a while. You just inspired me to do it.
October 2011
2 posts
September 2011
4 posts
Introducing G-Male
Video:
In a cross between Google Plus and Gmail, Comedivia brings the world G-Male, the perfect male boyfriend by Google. He knows all your interests and you will never forget anything again with him around, plus he is a really good listener. I am sure he will get along good with the Gmail Man.
Artificial intelligence at its best! G-Male is the project of Google to create a humanoid personal assistant/partner that learns, predicts and does what you need. No kidding:)
On Evolution, Learning and Intelligence
There is a strong connection between evolution, learning and intelligence. In this short piece I’ll try to track this connection for biological organisms and project it to machine intelligence.
The concept of intelligence is tricky to define. What is clear, however, is that learning is the crucial feature of intelligence. There are hardly any examples of biological organisms that are considered intelligent but unable to learn. Even bacteria can learn. The question is why learning is so essential? An obvious answer is adaptation. The world is changing and in order to survive the appropriate modifications in behaviour have to be made. The same function is fulfilled by evolution but it takes much more time than a life of an individual organism. Nevertheless, it works for microorganisms with short lifespan. Larger creatures live longer and evolution is combined with learning to achieve the sufficient rate of adaptation.
What about machines? Do we need learning, evolution or both to achieve intelligence in machines? The answer is much less obvious than with biological organisms. Unlike nature, learning and evolution implemented on a computer can be used to achieve similar goals. There is no binding to the reproduction rate or lifespan of an organism, a computer can run through millions of generations to come up with the optimal design for a robot in seconds. Computer simulation is often sufficient so there is no need for building millions of physical robots from intermediate generations. Artificial evolution can alter the behaviour of a robot during it’s ‘life’. For example, evolutionary computations can be used to construct and modify plans for a robot to accomplish a task. The same goes for machine learning, which may have other effects than it is possible in nature. For example, a robot may learn to modify its physical body according to the task at hand. Another interesting example which can be easily implemented on a computer but impossible in nature is Lamarckian inheritance i.e. learned skills and knowledge acquired during lifetime transferred to offspring. Intuitively, this kind of inheritance might be very desirable for evolving highly intelligent machines.
Modern intelligent machines such as well-known question answering system Watson and chess computer Deep Blue although regarded as intelligent are far from being biologically plausible. They utilize machine learning and evolutionary computations merely as tools to accomplish specific subtasks. Until now this approach was arguably more successful than attempting to copy nature and I tend to believe that it will stay this way.
You are welcome to express your opinion on the topic in comments.
August 2011
18 posts
Official Machine Learning Class from Stanford →
ml-class.com
on top of that ai class stanford will be offering a machine learning class
Stanford is doing a great thing here. Machine learning (ML), in many respects, is the core of artificial intelligence technology nowadays. Video of lectures from Stanford Machine Learning course have been avaliable on youtube for some time now. I watched the whole course and must say that the quality of the material is amazing. It provides a great insight into the field and Andrew Ng is a great teacher. There are reasons to believe that this new ML course will be even better since it is adapted for online usage.
Play
“Every now and then I try to read a novel but to me all novels are the same, they’re about a bunch of not very interesting people who make some mistake and their life gets very complicated and at the end they either find a way to… to get better or they get killed or whatever.”
—Marvin Minsky (one of the pioneers of the field of Artificial Intelligence) about non-science fiction literature
Robot cars - provably uncrashable? →
i-programmer.info
Autonomous driving is a hot topic nowadays.
- Companies such as VW release autopilot technology (video here).
- Google have demonstrated their driverless car (TED talk).
- Recent news about China building and testing its own robot car (see here).
The article linked in this post provides a brief overview of the technology behind autonomous driving.
