Andrew Cullen
adrcullen.bsky.social
Andrew Cullen
@adrcullen.bsky.social
AI (security, privacy, HPC) at University of Melbourne, researcher who rides bikes to go nowhere, and lifts heavy things for fun. Surprisingly interested in Iranian brickwork (just weird like that).
Good to see politicians going so quickly from responding to groundswell to trying to put forth action. But long term this isn't about CSIRO, it should be about how science is treated (and funded) in Australia.
November 25, 2025 at 10:13 PM
Reposted by Andrew Cullen
Sums it up.
November 19, 2025 at 10:49 PM
"CSIRO adding it would be looking for between $80m and $135m each year to renovate its ageing property portfolio."

Here I was thinking it was that humans that did science, not buildings. Yet one more set of cuts to finally yield the"sharpened research focus that capitalises on our unique strengths"
November 18, 2025 at 6:51 AM
On one level, I'm surprised that people are surprised by this - Chinese academics coming to American AI conferences has apparently become a near impossibility, especially for PhD students. On the other, it's still deeply concerning to see things potentially getting worse.
“The prohibited activities would include joint research, co-authorship on papers, and advising a foreign graduate student or postdoctoral fellow. The language is retroactive, meaning any interactions during the previous 5 years could make a scientist ineligible for future federal funding.”
U.S. Congress considers sweeping ban on Chinese collaborations
Researchers speak out against proposal that would bar funding for U.S. scientists working with Chinese partners or training Chinese students
www.science.org
November 15, 2025 at 10:46 AM
Reposted by Andrew Cullen
arXiv will no longer accept review articles and position papers unless they have been accepted at a journal or a conference and complete successful peer review.

This is due to being overwhelmed by a hundreds of AI generated papers a month.

Yet another open submission process killed by LLMs.
Attention Authors: Updated Practice for Review Articles and Position Papers in arXiv CS Category – arXiv blog
blog.arxiv.org
November 1, 2025 at 5:28 PM
Unwarranted detention, brought to you by the false positive rate.

Have reviewed AI papers proposing similar work, that haven't been flagged for ethics reviews. For authors working in adjacent spaces - this is how your work will be used. Reviewers - think broadly about potential harm. ACs - filter!
New: Videos show ICE/CBP agents are scanning peoples' faces on the street to verify citizenship. ICE has tool to instantly look up unprecedented number of databases with just a photo

“I’m an American citizen so leave me alone”

“Alright, we just got to verify that”

www.404media.co/ice-and-cbp-...
ICE and CBP Agents Are Scanning Peoples’ Faces on the Street To Verify Citizenship
Videos on social media show officers from ICE and CBP using facial recognition technology on people in the field. One expert described the practice as “pure dystopian creep.”
www.404media.co
October 29, 2025 at 9:15 PM
Reposted by Andrew Cullen
Translation for last year PhD students: You want to watch the ARC webpage tomorrow to see who will be looking for postdocs in 2026
ARC says Discovery Projects outcomes will be tomorrow (Tuesday). Linkage Projects (2025, round 1) on Wed. Over past ~year, it's often been at about 11am (Canberra).

My bot will pick up the change to RMS & post immediately.

ARC should email outcomes to lead CIs, but might take an 1hr or so for DPs
October 27, 2025 at 12:08 AM
This seems to be...interestingly timed, given the lawsuit regarding Adam Raine. From a security perspective, I'm curious as to what exactly OpenAI has done that leaves them confident about amping up how much "like a friend" LLMs will be, given our evolving understanding of the human risk landscape.
October 15, 2025 at 6:24 AM
Huge repositories of private information? Check. No information security at all? Check.

Rental applications are an absolute goldmine of private information - up to and including bank account information, passport numbers, job history and more, especially for foreign nationals.
September 18, 2025 at 11:02 AM
In today's installment of bad graphs and data: the power of setting the boundaries to fit personal views. What does accept mean? How does accept differ from undecided? There may have been a justification by the pollster, but that's not clear in the presentation.
Incredible stuff here from the Nine papers to combine “I don’t have strong feelings but I accept the fact we have purchased these submarines” with “support”.
September 16, 2025 at 4:10 AM
Reposted by Andrew Cullen
The National Climate Risk Report - my life for two years - has just been released.
www.acs.gov.au/pages/nation...
Minister Bowen press conference at: www.youtube.com/live/GLvd_fu...
My specific report on the weather and climate hazards: climateservice.maps.arcgis.com/sharing/rest...
September 15, 2025 at 12:28 AM
Hard to be proud of the institution over this. An organization that can pursue the Fisherman's Bend folly for years, and yet $110k to honor its commitment to an Australian intellectual institution is suddenly too much? And to just shutter it without even trying for alternate pathways?
It’s ludicrous and insulting to suggest that the richest university in Australia can’t afford the two part-time wages to run Meanjin.
Also: did they even try to save it? Donation drive? Philanthropic efforts? Offering it to another institution?
No.
The university council shut it down intentionally.
Exclusive: Australia’s second-oldest literary journal, Meanjin, is being shut down today for "financial" reasons, Crikey can reveal.
September 4, 2025 at 10:41 AM
This is such a simple thing, and would be so impactful. I've had part time PhD students who've had to make difficult decisions when juggling childcare, finances, and just surviving. This wouldn't fix it all, but would 100% move the needle.
My petition to the 🇦🇺 Australian government: make part-time PhD students' stipends tax exempt!

📋 Read and sign here: www.aph.gov.au/e-petitions/...
⏰ Deadline: October 1
e-petitions
e-petitions
www.aph.gov.au
September 3, 2025 at 8:09 AM
There's an argument that students don't want lectures. And as a lecturer, it can be demoralising teaching to a room of 6 out of 60. But a few of us at UniMelb have been experimenting with empowered learning. We've had almost 0 change in lecture attendance so far.

www.theguardian.com/australia-ne...
Adelaide University students and staff to stage sit-in over ‘travesty’ of fewer in-person lectures
Students unhappy about lack of course choice and flexibility, as they await clarification about which subjects will move online
www.theguardian.com
August 24, 2025 at 8:13 AM
Reposted by Andrew Cullen
Relaxation tips:
- put down your phone
- breathe deeply
- craft effigies of your enemies out of bones & hair
- bury these effigies deep in the misty woods
- summon forgotten gods and listen to stories of the old days, when moss & magic ruled the earth
- have some tea
August 12, 2025 at 2:25 AM
To anyone learning how to present data - study this graph from Gotham Polling (gothampolling.com/study-nyc-ma...). Study it in great detail. And then do absolutely everything differently.
August 18, 2025 at 10:15 PM
And it's not just USENIX - arxiv.org/html/2412.07... has an author being an accepted author of 80 publications in a single year, and authors with 15+ accepted submissions at NeuRIPS in recent years. You can't contribute to 80+ papers a year, and you definitely can't handle the associated review load
August 13, 2025 at 10:27 AM
Still remember at Uni trying to work out how long it was until 4:45pm. 6 maths honours students in the room (most ended up with PhDs). 6 different answers. All were wrong.

That we'd all been in the office for 48 hours straight at that point probably explains it. We would get blueberry wrong too.
In fairness, if you asked a bunch of PhDs how many B’s in blueberry, you’d probably get some wacky answers too
"PhD-level experts in your back pocket" is a completely nonsensical description of AI but a pretty good description of social media if you follow the right people
August 12, 2025 at 10:14 AM
The social license of AI Scribes ( ia.acs.org.au/article/2025... ) may only exist because users don't understand that their personal information doesn't just live in their GPs computer - it goes overseas. Also, having audited one in the past, I can also say that they can be dangerously rubbish.
Kobi refused a doctor's AI. She was told to go elsewhere
Unregulated AI scribes raising privacy, security concerns.
ia.acs.org.au
August 12, 2025 at 10:06 AM
Reposted by Andrew Cullen
"Recognition without sanctions is an empty gesture. It will not stop the bombs, the siege, or the starvation."
A very good statement from the Palestine Action Group Sydney on the Australian government's imminent recognition of a Palestinian state. As they say "We will not settle for symbolic diplomacy while genocide continues".
August 11, 2025 at 3:33 AM
Reposted by Andrew Cullen
A milestone for open AI: ETH Zurich and EPFL will release a fully open multilingual language model, trained on CSCS’s “Alps” supercomputer. Built for the public good, this model promotes transparency, inclusion, and innovation.

Read more:
A language model built for the public good
ETH Zurich and EPFL will release a large language model (LLM) developed on public infrastructure. Trained on the “Alps” supercomputer at the Swiss National Supercomputing Centre (CSCS), the new LLM ma...
ethz.ch
July 9, 2025 at 6:42 AM
Terry Tao makes a great point on analysing the properties of AI for performance in things like the Maths Olympiad. These claims are being treated as a sign of AI superiority, but rules and details are crucial for determining if this is all just puffery.

mathstodon.xyz/@tao/1148814...
Terence Tao (@[email protected])
It is tempting to view the capability of current AI technology as a singular quantity: either a given task X is within the ability of current tools, or it is not. However, there is in fact a very wid...
mathstodon.xyz
August 3, 2025 at 11:54 PM
Reposted by Andrew Cullen
We have open-sourced anonymized data and core analysis code for our developer productivity RCT.

The paper is also live on arXiv, with two new sections: One discussing alternative uncertainty estimation methods, and a new 'bias from developer recruitment' factor that has unclear effect on slowdown.
July 30, 2025 at 8:10 PM
Have seen a few news articles about this, but none seemed to include actual links - @sydney.edu.au is running a women only recruitment campaign for Engineering. The UoS page describing the program is www.sydney.edu.au/news-opinion... , with a link to the recruitment campaign at the end of the page.
July 28, 2025 at 1:53 AM
Reposted by Andrew Cullen
;)
July 15, 2025 at 6:42 PM