from the urls-we-dig-up dept
Machine learning is a fascinating field that is increasingly becoming a part of more and more aspects of our lives. Not too far in the future, we’re supposed to have all kinds of robot helpers — driving cars for us and doing all the boring jobs. We’re not quite there yet, but it may only be a matter of time before it’s common for a robot to ask, “Do you want fries with that?” Here are just a few interesting examples of machines learning how to be more like us.
- ConceptNet 4, a version of an artificial intelligence developed at MIT, has scored about as well as a human 4 year old on an IQ assessment designed for young children. ConceptNet is programmed to make some common-sense associations, but it doesn’t do well on answering “why” questions. At least we’ve gotten AI past the terrible two’s…. [url]
- Neural networks are one method for modeling how biological brains and nervous systems work, and these simulations are getting better at mimicking how actual brains learn. Increasingly complex Boltzmann machines are learning to perform activities similarly to how a person would — and making some of the same mistakes that people do, too. [url]
- Programming computers to use our spoken language is difficult because people often use the same word to convey very different meanings, and dictionaries aren’t much help to a computer for making clear distinctions between these meanings. Computational models poring over vast amounts of unstructured data on words can generate more computer-friendly dictionaries that map word meanings into groups. The result could be automated paraphrasing programs that can re-state a sentence without completely changing its meaning. [url]
- A startup called Vicarious is developing software that tries to process visual information like human brains do. Building a better neural network algorithm could create commercial software that can do things like help diagnose medical images. Expert systems were supposed to replace doctors decades ago, but maybe the technology has finally caught up with the science fiction predictions? [url]
If you’d like to read more awesome and interesting stuff, check out this unrelated (but not entirely random!) Techdirt post via StumbleUpon.