Everyone talks about artificial intelligence revolutionizing the world as we know it. Yet all we seem to see are fairly basic bots: an Alexa that only understands half of what you say and computer vision applications that can’t tell basic things apart .
Why is this?
To answer this question, let’s look at what makes machine learning different from software. The latter is—in essence—a set of very clever rules, that exist independent of the inputs, written on top of assemblers and compilers. None of these features exist for ML: What matters is the training data, and the quality of how it is structured and labeled. The labelling of the data is the programming, and right now it is the single biggest bottleneck to making AI deliver on its promise.
04:30PM - Day 1