based on
Protips from Vern Bengston, USC
Dan Ryan
9.21
SQ3R+25W
S
Q
R
R
R
25
Z
urvey
uery
ead
ecite
eview
words
otero
CONTENTS
Reading scholarly work is different from reading a novel or a blog post
S
Q
R
R
R
25
Z
Survey
ARTICLE: Assess the size. Realistically estimate reading time. Read abstract and first paragraph. Skim conclusion. Flip through. Write out section headings if appropriate. Sketch a mind map?
5 minutes
STOP
BOOK: Assess the size. Study the table of contents. Skim introduction and conclusion, first chapter and last. Flip through book. Skim the index.
~350-400 words/page
330 pages
200-250 words/minute
8-11 hours reading time
115,000-180,000 words
Survey
Survey
Google tool produces funky word meaning results.
Using AI for criminal justice stuff. Pro publica scandal.
Weird AI video game result
machine learning and human values
both a technical and a social problem
Survey
Look over the Conclusion
We must take caution that we do not find ourselves in a world where our systems do not allow what they cannot imagine - where they, in effect, enforce the limits of their own understanding.(327)
Survey
10 pages
~900 words/page
9000 words
150-175 words/minute
50-60 minutes reading time
Survey
Survey
Survey
Query
What major point (or finding) is here?
2 minutes
What are the major questions that the book/article wants to answer?
DON'T refer back to manuscript. DO focus on "first impressions"
What do I want from this?
STOP
You May Ask yourself...
WRITE IT DOWN.
The Alignment Problem
Book explains state of the field, what CS people are thinking about and what "critics" are thinking about. I can get an overview and sense of where the field is headed and I think come to understand some technical details too.
Zombie Algorithms
It's a philosophy article arguing in that style. Basic point: algorithms are not like people and so cannot be morally responsible. This means we have to make their creators or owners accountable. I will learn some philosophical concepts and how philosophers are participating in conversations about AI ethics.
Query
Read
Perhaps eschew highlighters.
Annotate
Underline. Write in margins.
High level outline or mindmap.
the hell out of everything you read
Recite
Do I have questions?
2 minutes
STOP
Three things I should remember
Major points made by author
Close book/article
Do I have criticisms?
Review
Did your recite get it right?
Questions answered? Criticism still stand?
Look back over text and notes.
STOP
5 minutes
25 Words
Write a 25 word summary of what the piece says and why it matters for this project or assignment.
Extra: draw the argument
5 minutes
5 minutes
25 Words
Where does "responsible" in responsible AI "land"? Moral zombie concept helps us see that until we invent machines that can be sentient, the humans behind the algorithms are accountable.
AI alignment means "values" of machines are consistent with human values. Lots happening recently and lots of problems still to solve.
NEVER
JUST PLOW THROUGH FROM
FIRST WORD TO LAST
One more thing
zotero
https://www.zotero.org/
deck
By Dan Ryan
deck
- 126