DRI Unconference

Spread out in the foyer area, the bar and the prayer room as well as in here

Guiding Question:

Guiding Question:
What developments in our field(s) do you think are:
(a) Most threatening
(b) Most exciting
that are coming our way in the next 3 years?
How can we deal with them?

Pick a group below and jot down some thoughts!

Online group 1

  • Following up from the GDC Keynote, AI is an obvious threat. Jensen touted the idea of asking AI to build the whole workflow. But it's also an exciting opportunity. What new things we could we do?
    • AI+robotics as advertised might even be an issue - no need for someone in the data centre if the robot can get in and do things instead.
  • People are going to be an issue. Both from AI doing things formerly only humans could, but also because lots of critical research computing teams are very small. The 'bus factor' is concerning in a lot of cases - not that folks are likely to be hit by a bus but that people might want to move and do something new. Difficulty with recruitment due to salary differences, you really have to want to work at the university. But it's not all doom and gloom. As we get further in the future more and more technical people who made their careers on the back of research computing teams are going to get into management positions and the value of RC is only going to increase. At the same time, students are also coming up the pipeline in the next few years to research computing roles where they've lived with technology in ways we never have and it's exciting what that experience will offer.
  • Storage is a threat. There's so much data to keep, both to satisfy funding requirements, but also for more practical purposes. It should be more cost effective to keep a hard drive spinning with a few TB of data that might be needed (while satisfying funding requirements) than it would be to regenerate that data using compute resources that are already heavily in contention. We must have storage capacity.
  • Back to feeding the pipeline for new professionals, how we do this is a challenge. Particularly how do we get people from the social sciences/humanities involved? It was quite striking when most of the room raised their hand as being from a STEM background. What are we missing out from our English Language, Music, etc., majors? How can we get those people into research computing so that they see our teams as a viable career path?
    • We have to engage at an undergraduate level. But these are students who may have got into their field to avoid maths, so the programming has to have a purpose. Things like Processing for musicians or GIS are a way in, but leveraging this into research computing needs support from lecturers. Maybe even as little as just making students aware that what they're doing is research computing. Pulling off the Mission Impossible or Scooby-Doo masks to reveal the programming underneath. Or even earlier - there are TI N-spire calculators with Python on them.
  • Following up on the technology and teaching, ChatGPT is going to be an unfair means issue. Do we make things open book? Multiple choice? Online? In person? Anti-cheat software like is used in video games? Extra security like used in banking?

Online group 2

Online group 3

Group A

Met Office/UKCEH/STFC/UCL/Leeds/York (MUSULY)

Data and metadata standard divergence - threatening
More and more data being published but platforms creating their own schemas/allowing people to publish data with incomplete metadata - we do our best to publish data but with incomplete metadata it may never be findable/FAIR.
Trustworthiness of data - knowing how it was generated and trusting sources.
Different discilines may have set standards, but we want a way to bridge overlaping areas.
Not possible to have one standard for everything - in need of a central overlap
The deeper weve talked the more weve fond different domains do things differently
Interesting thoughts:

Consolidating training efforts - exciting
Weve discussed this wee that there is a lot of duplicated effort in terms of training across universities - exciting prospect that we could pool resources and streamline training across the field.

Group B

KCL/ICL/TUoS/Bristol/STFC/EPSRC

Threating developments

AI - threatening. Hype/perception of AI causing reduction of funding/research because "AI" can do it instead of people

Security - changes made to support the need for enhanced security from an enterprise perspective - we will likely see shifts towards securing infrastructure that may affect the way that DRI professionals work (or the flexibility that they have).

Personel - loss of staff, spread too thin, career paths. Funding for infra often outweighs money for personel to use the infra (URKI are working on this though!)

Exciting developments

TREs - Trusted research enviornments

GPUs - new developments, different providers offering new architectures, etc.

TPUs - interesting hybrid between GPU and ASIC - likely more developments in this space.

Cloud - azure classroom (budget control features!)

Federated access to DRI resources

Group C

Uncertainty of the future of funding
Increase in the fragility of systems and projects
Lack of people resources
Perks received previously by academics staff is being degraded
Hostile attitude from government towards the academic sector
The funelling of money towards commercial industry sector. How can we reroute this through academia?

Positive
Can push exciting science
The more exciting we are the more productive we are
Scale of computation available. Allow us to explore problems at new scales.
We now have the ability to run these problems with the computing power we have
Some of the applications of machine learning and AI that have become viable

Group D

Group E

Group F

Group G

Group H

Group I

Group J

Group K

Group L

Group M

Group N