Bayesian
Latest
Harvard researchers make better, smarter walking aids
Humans don't all look, talk, or walk the same, with us shifting our weight and style in order to save much energy as possible. This adaptability is a problem for researchers who want to build assistive devices for folks with mobility issues, however. It's what has prompted a team out of Harvard to craft an algorithm that can determine the wearer's individual needs after just twenty minutes of analysis.
AI is learning to speed read
As clever as machine learning is, there's one common problem: you frequently have to train the AI on thousands or even millions of examples to make it effective. What if you don't have weeks to spare? If Gamalon has its way, you could put AI to work almost immediately. The startup has unveiled a new technique, Bayesian Program Synthesis, that promises AI you can train with just a few samples. The approach uses probabilistic code to fill in gaps in its knowledge. If you show it very short and tall chairs, for example, it should figure out that there are many chair sizes in between. And importantly, it can tweak its own models as it goes along -- you don't need constant human oversight in case circumstances change.
An AI algorithm can draw letters as well as a human
Researchers claim to have made a breakthrough in artificial intelligence by giving machines cognitive powers similar to humans. The team from MIT, York University and the University of Toronto first trained an algorithm to draw characters in 50 languages by studying the required pen strokes. Once completed, it was able to successfully draw a new character that it had never seen before, meaning it had essentially "learned" the skill. That might not sound impressive, because we humans can do it easily. But so far, similar feats have only been done by large neural networks that require huge databases of images and learn more by brute force than smarts.
Friday Favorite: SpamSieve 2.76
My Friday favorite is SpamSieve. We have mentioned it a few times previously, but since it has recently been updated to version 2.76 I wanted to sing its praises again. It's the best way I've found to deal with spam. Using Bayesian filtering, SpamSieve installs as a plug-in to your mail client and lets you mark messages as spam. As you do, it builds a a corpus file of rules telling determining what is spam and what isn't. The more messages you mark, or train, the more accurate SpamSieve gets. I've been using it since November of 2003 and after years of training, it's so accurate that it rarely fails to catch an errant spam encrusted message. When it does, using either a keystroke sequence or a pulldown menu from your Mail client you can train it as spam. At the start, it's quite labor intensive since you have to mark a few hundred messages for it to really start working, but it pays dividends. After a while, you'll have a personalized set of inclusion/exclusion rules that gets better over time. To give you an idea, yesterday I received 307 emails. Out of those SpamSieve correctly marked and moved over 30 messages and missed only 2 that needed training. This is a shot of my corpus screen showing how many messages have been filtered and how many words were read resulting in messages being regarded as spam or good. Yes, over 15,000 messages is a big number, but by being cumulative, SpamSieve gets more and more accurate over time. SpamSieve allows you to import or export the corpus file so if you get a new computer, or decide to use a different email client, you lose nothing.
