At the AWS Summit in San Francisco on Wednesday Amazon Web Services invited a handful of tech typers to see a demonstration of AWS DeepLens, its forthcoming camera tuned for deep learning tasks.
Announced late last year and given away in limited quantities to select developers, the image capturing kit is intended as an AI onramp for coders.
Amazon has been taking pre-orders for the device since its announcement; it now claims the device will ship June 14, 2018, at a cost of $249.
"The goal here is for developers who have limited or no machine experience to use deep learning, to get up and running quickly," said Matt Wood, general manager of artificial intelligence for AWS.
There's a good reason for developers delve into AI disciplines: Companies are looking to hire AI experts. As with security, high-level talent is scarce.
But beyond potential employment benefits and irrepressible curiosity, there's not really a compelling reason to take up tinkering with machine learning unless you regularly deal with challenges involving large data sets or you really just cannot resist a shiny gadget.
For coders involved in finance or security, or social photo apps that handle millions of images, machine learning is a no-brainer. But for those without massive datasets – millions of user profiles spirited out of Facebook via its now-neutered API that have been squirreled away on hard drives, for example – it can be a struggle to come up with non-trivial use-cases.
Amazon last year suggested replicating the hot dog detection app featured in an episode of HBO's tech skewering comedy Silicon Valley. And indeed, DeepLens can do that.
But people do that even better. Yes, you could write an app to detect when your dog has climbed onto your couch. But you don't have to pay Amazon for that privilege; you could just pay attention.
Or you could take on a slightly greater challenge by assembling an image recognition system with a Raspberry Pi, Python, and OpenCV from scratch. It would be an educational experience in its own right.
But if you're already hooked on AWS cloud crack and can't wait to invite another IoT device into your life, the bill shouldn't be too high. Woods said most models can be trained within AWS' free tier. ®