Take a photo of a barcode or cover
ystasyuk 's review for:
AIQ: How People and Machines are Smarter Together
by Nick Polson
This was a quick and easy introduction to the state of artificial intelligence and data science today. Unlike modern media sensationalism and allusion to skynet, this shows a that most AI is mathematical modeling and predictions based on large data sets, its not about sentience or consciousness. There is math in this book, but its lucidly explained without resorting to pedantic mathematical explanations. There is a great explanation of Bayes theorem using searches for lost submarines. The key takeaway is that consulting prior probability and adjusting this in light of new evidence, helps us come up with more accurate conclusions.
This idea is repeated throughout the book, from stories about Issac Newton missing simple coin size averaging issie during his stint as treasurer to stories of missing kidney decline by failing to draw a trend line.
The assumptions you bring into artificial intelligence are pivotal to it's product. Another memorable anecdote is about an army tank identification algorithm. Thousands of images of an empty forest were compared to images containing hidden tanks. The algorithm performed perfectly on the sample dataset but failed when deployed in the real world. Why? Because all the forest images were taken on a cloudy day, while tank pictures were taken on a sunny day. Three are biases everywhere and our assumptions are key.
This idea is repeated throughout the book, from stories about Issac Newton missing simple coin size averaging issie during his stint as treasurer to stories of missing kidney decline by failing to draw a trend line.
The assumptions you bring into artificial intelligence are pivotal to it's product. Another memorable anecdote is about an army tank identification algorithm. Thousands of images of an empty forest were compared to images containing hidden tanks. The algorithm performed perfectly on the sample dataset but failed when deployed in the real world. Why? Because all the forest images were taken on a cloudy day, while tank pictures were taken on a sunny day. Three are biases everywhere and our assumptions are key.