bwagy, the blog of Ben Young

Advertising has the media rating council. The council seeks to validate that independent measurement does what it says on the tin through an audit process.

AI algorithms will have the same need.

Each is a black box and with this he need for independent ratings of algorithms will only rise.

Especially in the enterprise sector.

This uber self-driving accident is case in point, there is no way for anyone else to verify that indeed the stated cause is what caused it.

Think of it as a quality verifier. Who is building this? Is anyone yet?

Listen to these (scroll to the bottom).

You can see the real effect of how machine learning can round out the corners on a product or service.

Apple used it to remove the robotic effect Siri has and bring emotion in.

It’s something -> that without machine learning, was just simply put going to require too much effort (otherwise they would have done it already).

When you listen to the audio, imagine this metaphor applied to any software experience. It literally is just leaps and bounds ahead.

We’ve been playing around with Machine Learning, to help uncover patterns in content.

And there’s two significant components
1) Definition
2) How variables stack up against that definition

You can look at the elements, imagine you want to use machine learning to find out how to build a good car.

You would load up all the good cars, that within itself is a definition, what is good?

It could be design lead, which could be established through design awards, or design nominations, it could be sales, it could be publicity.

Something quantifiable.

Then you break down it down to detect all the variables that make up a good car, things like four wheels, the gradient of the curves, acceleration, weight, colour.

This helps you then understand what a good car looks like.

But good is subjective.

Just remember that, data analysis is great but it always comes back to what are we doing it for.