Research Grants
Key aim:
Create a roadmap for one or more
popular books on AI risks.
The vision behind
this project:
- Nearly everything useful that gets done, gets done because some set of people choose to do it.
- Peoples' mental landscapes affect their actions.
- More specifically, there are key concepts, metaphors, and cognitive skills that can make people more interested in reducing existential risks, and more able to take effective action to reduce such risks.
- Popular books can be a key vehicle for conveying such concepts, metaphors, and/or cognitive skills (e.g., Malcolm Gladwell's book Blink).
- For a book to become highly popular, it needs to offer content that the target audience is strongly motivated to gain, that is easy for the target audience to remember and to integrate into their world-view, and that is "remark-able" in the sense that target members, once exposed, will want to remark on it to their friends.
- By testing actual responses from the target audience to each of a wide range of possible book chapters and framings, we can increase the odds of writing a high impact, "remark-able" book.
- Therefore, by creating a book on existential risks in accordance with the principles of modern marketing, including the idea of testing potential book and concept variants, the Singularity Institute could significantly increase the number of people interested in reducing existential risks.
Proposed grant
activities to realize the popular book vision:
Step 1:
Aims.
Step 1a.
Generate a clear picture of our aims in writing popular books.
Deliverable:
A brief “what popular books can do” section that can be added to our SIAI strategic vision document. The section will include:
- What target groups we would like to influence in what ways;
- How exactly each desired aim (in terms of target group and desired effect) might be expected to reduce existential risks.
December 3
Step 1b.
Bounce the
results of Step 1a off each of Peter Platzer, Michael Vassar, Eliezer
Yudkowsky, Carl
Shulman, Anna Salamon, and Michael Anissimov, and off of interested
outside players such as Toby Ord, Nick Bostrom and Gaverick Matheny.
Deliverable:
Record of disagreements and
modifications
to the sections as appropriate.
Delivery date:
December 19
Step 2: Gather
content.
Brainstorm concepts and ideas that can be developed into book sections
advancing the purposes identified in Step 1. Pay special
attention to
existing SIAI content that is currently only on the web - for
example
Overcoming Bias posts, old email discussions from the mailing list SL4
and to ideas that have been floating around SIAI but have never been
written up.
Deliverable:
For each target group and
direction
in which we wish to influence that group identified in Step 1, a list
of relevant concepts, thinking skills, chapter headings, etc. that we
may wish to include. The list will be inclusive and should contain
every
concept worth market testing. I.e., it will include
everything
that we think might plausibly be worth including in
the book.
Delivery date:
December 26
Step 3: Streamline
content.
Turn the content from Step 2 into memorable phrases and frames.
Deliverable:
A list of Google AdWords ads
to use in market research.
Delivery date:
January 2
Step 4: Test target audience responses.
Do market and viral
marketing/memetic (what people will be more likely to remark
on to their
friends and associates) research. Gather empirical data as to which
of our possible chapter titles, book titles, and book framings are of
most interest to the target groups.
Attempt this by counting
click-throughs
for Google Adwords using different phrases and promises; if the Adwords
testing method is inviable for whatever reason, conduct some other poll
of the target group.
Deliverable:
Empirical data, added to the list
from Step 3, as to how appealing our various chapters, book titles,
and back-cover book descriptions are to the target group(s).
Delivery date:
January 23
Step 5: Make
integrated book
plans. integrating (i) market demand concerns, (ii)
book
flow, size and unity concerns (iii) impact on readers' understanding
of existential risk (iv) effort required, and what our group can
reasonably
accomplish given time and resource constraints.
Deliverable:
Integrated book plan
document,
possibly also a skeleton of the book including chapters, section
headings
and outlines of content.
Delivery date:
February 6
Step 6: Set up for
next
steps.
Step 6a.
Set up for
those steps that are not covered by this grant.
Deliverable:
A list of intermediate pieces of content (journal papers, web documents, etc. that we can write) that are both:
- immediately useful in themselves, and
- useful steps towards a strong popular book or books.
Intermediate pieces of content may include:
- Academic journal articles that serve as chapter drafts or key references for chapters;
- Blog posts or post sequences that enable rough drafts of chapters to be subjected to market testing and reader feedback during development;
- YouTube videos or conference talks that can be useful as further market testing, and as rough chapter drafts.
Delivery date:
February 20
Step 6b.
Write a proposal for each such to-do item in roughly this grant format (though more briefly if the todo is small). For each item, specify:
- Resources required to complete the project (what skills are needed, for how many hours; whether any resources other than SIAI staff time are needed);
- How the project would be immediately useful toward reducing existential risks;
- How
the project would
facilitate the later creation of a popular book.
Deliverable:
A collection of grant proposals.
Delivery date:
February 20
Step 7: Follow-up
report.
Step 7a.
Record actual
required inputs (time, labor, funds) and compare to expected and
budgeted
amounts. Post to the SIAI website.
Deliverable:
"Actual budget" follow-up document for the SIAI Peter Platzer popular book project
Delivery date:
February 20
Step 7b.
One and two years after the completion of
this project, briefly reconsider the concept of a popular book, the
degree to which the "next steps" from step 6 have been completed, and
any additional projects which may make sense.
Deliverable:
Follow-up reports which discuss:
- the current state of the book(s)
- which of the proposals from Step 5b have been begun, and which have been completed
- any
next steps which may make sense.
Delivery dates:
February 19,
2011, February 18, 2012
Budget:
Total budget:
$2,500
Budget breakdown:
$300 for Google AdWords purchases,
for carrying out other survey methods if AdWords fail (e.g.,
use standard survey methods to get students at a local university, for
example Stanford, to complete surveys), or other materials.
$1,000 of SIAI staff time for
high-level
planning and content generation and to supervise and guide task
outcomes,
working out to 50 hours at $20/hour. (Core SIAI staff can earn more
than this, but are prepared to work on existential risk mitigation for
less than they can make elsewhere)
$1,200 of SIAI research assistant
time, working out to 120 hours at $10/hour. (Again, the
market value
of research assistants' time would be higher than this, but SIAI has
a backlog of skilled potential hires who are willing to work on
existential
risk reduction for less than they can make elsewhere.)
How this grant,
if funded, can be expected to reduce existential risk:
Benefit 1.
The grant will increase the
odds that we write well-planned, high-impact books in the relatively
near future.
Benefit 2.
The grant will also increase the
odds that our smaller, nearer-term projects (journal articles,
conference
talks, popular blog posts, videos) are of high interest to target
groups.
Cost 1:
Potential diversion
of SIAI staff time and focus from other projects. (Ideally, the labor
costs are covered by the grant money; but if the work takes longer than
is budgeted for, this assumption may be false. Actual costs
should be
assessed, and written up, after the project work is completed. Also,
although SIAI staff work for $20/hour, the opportunity costs of their
labor is greater, since the number of people able and willing to work
at this skill level for $20/hour is in limited supply.)
Why these benefits matter: Right now, the core of people working on existential risks is far smaller than one might like; if we can increase this core, humanity's odds improve. Benefits 1 and 2 can begin to increase this core.