The Era of the Algorithm

The future should not be self-driving. I think the best way we can put these fears and anxieties aside is to take the driver seat. Be intentional about this; be kind to one another.

This old song that if you're free you're not the customer, you're the product. It's quickly becoming this: You are the training data. By using this system, you are giving up information to some purpose that you may not be completely aware of.

What I found the more that I work with and design these machine learning experiences is that it's less about designing for a fixed path through information and much more about trying to put some guardrails up around the weird stuff that the people will ask these 'smart systems.'

The important thing to understand about machine learning is that what they're great at is figuring out what's normal. And then predicting the next normal thing. . . . Question is, what if normal is garbage? What if we're giving it data that has inherent bias in it?

How do we avoid building our ugly past into the future? Maybe put another way: How do we think about gathering data that reflects the future we want to build?

We need to get people on our teams that reflect the broader community.

It's really about setting appropriate expectations for what the system can do and channeling behavior in the right way.

A lot of the times we think about machine learning, algorithms, as being the work of data scientists and of engineers. . . . There is also an important role in design for all of it, for the presentation of this information.

If we're taking all of the past data to predict the future, doesn't that mean we're going to have all of the problems of the past?