Name: Responsible AI with TensorFlow (TF Dev Summit '20)
Uploaded: 2020-04-04T05:06:50.000Z
Duration: 21 min 52 s
Description: Thousands of YouTube videos with English-Chinese subtitles! Now you can learn to understand native speakers, expand your vocabulary, and improve your pronunciation...

and I'm here to talk to you about responsible AI

I'll focus on fairness for the first half of the presentation,

Today, I'm here to talk about three things.

And how does it affect an ML system exactly?

call throughout the presentation a fairness workflow.

Surprisingly, this isn't too different from what

for example, a debugging or a model evaluation workflow.

Finally, we'll introduce tools in the TensorFlow ecosystem,

that enables easy evaluation of commonly used fairness

We must acknowledge that humans are at the center of technology

design, in addition to being impacted by it,

and humans have not always made product design decisions

that are in line with the needs of everyone.

Quick, Draw! was developed through the Google AI

experiments program where people drew little pictures of shoes

Most people drew shoes that look like the one on the top right,

so as more people interacted with the game,

the model stopped being able to recognize shoes

is then amplified by fundamental properties of ML--

aggregation and using existing patterns to make decisions.

Minor repercussions in a faulty shoe classification product,

perhaps, but let's look at another example that can

to protect voices in online conversations

After its initial release, users experimented

with the web interface found something interesting.

The user tested two clearly non-toxic sentences

that were essentially the same, but with the identity term

Only the sentence using gay was perceived by the system

as likely to be toxic, with the classification score of 0.86.

When used in practice, such as a content moderation system,

this can lead to the systematic silencing of voices

at any point in the ML pipeline, from data collection

and handling to model training to deployment.

bias primarily resulted from a lack of diverse training data--

and in the second case, examples of comments containing gay

However, the causes and effects of bias are rarely isolated.

It is important to evaluate for bias at each step.

You define the problem the machine learning

oftentimes checking, analyzing, and validate it.

You build your model and train it of the data

And if you're applying ML to a real world use case,

your model, as we'll see throughout the next few slides.

The first question is, how can we do this?

isn't that different from a general model quality workflow.

The next few slides will highlight the touch points

where fairness considerations are especially important.

Consider what your metrics and fairness-specific metrics

are actually measuring and how they relate to areas

to evaluate on should be carefully selected

Subtitles ListPlay Video

Responsible AI with TensorFlow (TF Dev Summit '20)

recognize

positive

bias

identify