Academia
# The ‘rat race’ of Computer Science Academia
There is no mechanism today, other than time, donations, and personal social platforms, for researchers to support other researchers’ work. Every act of support is out of selflessness and there is a lack of incentive for cross collaboration other than having your name on another paper. The reward system in this community is highly dependent on your ability to make your research well known and marketed. Shrey Jain
Source: Just ask for Generalization
My guess is that the research community tends to reward narratives that increase intellectual complexity and argue that “we need better algorithms”. People pay lip service to “simple ideas” but few are willing to truly pursue simplicity to its limit and simply scale up existing ideas.
The typical research project might last 4 years at the longest because that is the max duration of a PhD and post-doc. An average PhD candidate wants their own unique project; they don’t want to continue someone else’s project. So the academic world is a rapid succession of short-lived projects1
# Tunnel Vision
Source: Mimetic by Brian Timar, see also mimetic
“Graduate programs select for intensely competitive individuals with highly specific skills, often with negligible market value outside of universities. A strong desire for publications on esoteric topics is inherited from senior postdocs and professors, making tunnel vision especially acute.”
# Incentive Structure
Academia feels more pure, more playful, than industry? More of a ‘ constructionist’ approach, freedom to ask your own questions
Can we create the energy of DARPA outside of government funding? A modern day Xerox Parc or Bell Labs? A research institutions perhaps?
More on incentives
# Research at Microsoft
Session w/ Jim Pinkelman, Ph.D. Been at MS for 19 years, 11 years in Microsoft Research
Most of MS research is not necessarily driven by a technical challenge that a product group faces. Very driven by research interests of new hires.
Missions
- Advance the state-of-the-art in CS
- Rapidly transfer technologies to Microsoft products and services
- Incubate disruptive technologies and new business
95% of internal research is peer-reviewed + published
1990 memo
- “We have some unique potential to productize research”
- “We should still listen to external research, but in some areas there is a big benefit in owning the tech”
Horizons
- Horizon 1: Near term, ~6mo-1yr, what technical challenges do we need to overcome?
- Horizon 2: Medium term, 1-4yrs, idea of general challenge needed to be solved
- Horizon 3: Long term, 5-10yrs, advancement in the field
MSR sub-groups
- MSR Labs: Long term basic research
- MSR NExT: New experiences and technologies
-
From Stephen Fay: “I don’t really understand where this claim comes from. In physics there are many huge international collaborations spanning decades (e.g. large radio arrays in poles+south africa+Canada, LIGO + LISA, CERN, James Webb, building quantum computers is going to take large collaborations even if research goes underground). I can’t really speak to other fields but it would be nice to have a bit more context to this claim.” I’ve now clarified this critique to mostly target computer science/theoretical fields. I think any research field with sufficient barrier entry/requirements for access to infrastructure/hardware/physical resources does necessarily require pooling of resources on an institutional and often multi-year/decade long timespan ↩︎