No algorithm is perfect. Sometimes, emails we’re expecting end up in our spam folders. Sometimes, online map tools can give us bad directions. Sometimes, online dating sites match us with creepy, sullen weirdos. But if you were around for the salad days of the internet, when that guy at work had a Geocities site with the dancing baby .gif and you were searching for stuff on Altavista, you’re aware of how much better those algorithms are. Some of that is that we’ve improved through trial and error; we’re simply building smarter machines. But a sizable part of the improvement is something called human-in-the-loop learning.
The basic principle is simple: no matter how precise or advanced a machine algorithm is, humans can often make it smarter. For example, we parse language better; we recognize sounds and images more accurate. To put it another way: algorithms have a tough time with sarcasm or figuring out whether an image happens to be explicit. Humans simply don’t.
It’s through this lens we’d like to share with you a free report we partnered with Ted Cuzzillo and O’Reilly Media on. It’s called Real-World Active Learning, Applications and Strategies for Human-in-the-Loop Machine Learning. You can download it over on our site to learn how crowds of people are working to improve everything from job postings to map accuracy to clothing recommendations. It covers when active learning works best, some basic principles around human-in-the-loop learning, how to manage a crowd to improve your data, and plenty more.
If you’re interested in how machines get smarter working in tandem with the crowd or how you can use a crowd to work on problems machines and algorithms can’t quite handle on their own, we think you’ll enjoy it.