Lesson 5 - Official Topic

[rachel (Rachel Thomas)]

This is a wiki post - feel free to edit to add links from the lesson or other useful info.

Note that the video and this thread is called “lesson 5” because that’s what this will be in the MOOC, even although it’s week 4 of the course

Lesson resources

• Edited video (private)
• fastbook chapter 3
• Non-beginner discussion

Links from lesson

• Rachel’s longer lesson on disinformation & coronavirus

• Rachel’s follow-up post on privacy & surveillance during pandemic

Other useful links

• Datasheets for datasets
• Weapons of math destruction
• AI Generated Faces
• Papers/repos/tools on how to check for bias in your datasets/algorithms AND address it
• Markkula Center - Ethical Toolkit
• The Pollyannish Assumption
• Understanding the context and consequences of pre-trial detention
• Fast.ai community ethics resources & discussion thread. Author credit: nbharatula
• Montreal AI Ethics Institute Weekly Newsletter #6: Radioactive data, attacking deep RL, steering AI progress, sucker’s list, AI ethics in marketing and more …

[username_not_found (unknown)]

Best part of the week.

[Raymond-Wu]

Just wanted to ask here because there have been several questions in the official thread with no response:

Where did the official study group Zoom go? The link is no longer functional.

Edit: Also super pumped for the ethics lesson!

[ilovescience (Tanishq)]

Will chapter 4 be covered? I don’t think we finished it last lesson.

[jeremy (Jeremy Howard)]

@Raymond-Wu

Where did the official study group Zoom go? The link is no longer functional.

Some enterprising person will need to replace it with something else for now - I’ve been flat out on masks!

[hiromi (Hiromi Suenaga)]

mic 1 is quieter

[Farah (Focal)]

Very important lesson, everyone in the industry show watch this lecture.

[Albertotono]

Awesome topic, I am delighted.
How can we break this feedback loop? Can we use a “surprise effect” on the recommendations? How can we implement it?

[jeremy (Jeremy Howard)]

@Albertotono

How can we break this feedback loop? Can we use a “surprise effect” on the recommendations? How can we implement it?

Human in the loop is critical.

[init_27 (Sanyam Bhutani)]

On masks, I wanted to add 1 thing that I just recalled, Even batman told us:

[Raymond-Wu]

What sort of pre planning or pre deploy steps can a team take when deploying a model to avoid things like a feedback loop?

[nchukaobah (nkiruka chuka-obah)]

Quick google search with Nigerian names does not have the same poor results as Sweeney Latanya. Interesting observation.

[jeremy (Jeremy Howard)]

Can you give me an example name? I wonder if it depends on location.

[pinaki]

Are there references re IBM and Nazis and their complicity etc. Didnt see any in the video.

[nchukaobah (nkiruka chuka-obah)]

Chika ejebe, Chinedu Ojokwu, Ikenna Uchenna, Ikechukwu chizoba. Location, like you said, could be an issue. Also, these names are for a particular tribe in Nigeria. Other tribes may be different.

[nareshr8 (Naresh)]

He is the one who covered parts that need not be covered and left open parts that were to be covered …

[gamino (Ariel Gamiño)]

There’s a book about it “IBM and the holocaust”

[jeremy (Jeremy Howard)]

@pinaki

Are there references re IBM and Nazis and their complicity etc. Didnt see any in the video.

IBM & "Death's Calculator"

[ilovescience (Tanishq)]

There is a wikipedia page on this with references:

IBM during World War II