Toby Murray
@tobycmurray
Followers
1K
Following
7K
Media
180
Statuses
5K
Go find me at @tobycmurray.bsky.social. Assoc Prof at @UniMelb and @cis_unimelb cyber lead; @DSI_VIC Director; @UniofOxford DPhil (@HertfordCollege); He/him
Melbourne
Joined February 2013
If data is the oil of the new economy, then we must be living in the age of the everyday environmental disaster.
3
18
40
A cautionary tale also with relevance to those writing LLM powered “agents”.
Letting chatbots run robots ends as badly as you'd expect via @theregister.
0
0
1
If you want to understand why “predict when people will do crimes” is utter snake oil, check out @random_walker and @sayashk’s excellent book “AI Snake Oil”. Along the way you’ll learn also why not all AI is snake oil and how to sort the wheat from the chaff. Highly recommended.
They already have our medical records, now Palantir ( run by Peter Thiel) want to get into the predict-when-people-might-commit-a-crime business. Thiel’s protege is JD Vance. Apart from that, have a lovely Saturday.
0
0
4
So much fuss about AI “finally” aiding scientific discovery. But “AI” has been helping humans “discover” theorems and proofs for decades.
0
0
4
“Did I know Kurt Cobain was a drug addict? Anthony Kiedis? Trent Reznor?. I think so, but I'm not sure. I was too busy sorting those CD singles in alphabetical order.” Not sure than teenage me has ever felt more seen.
0
0
0
Minimum age of criminal responsibility in Australia: 10 years .ACT: 12.NSW: 10.NT: 12 (for now).QLD: 10.SA: 10.TAS: 10.VICE: 10.WA: 10. Proposed minimum age to legally use social media: 16.
No SCREEN TIME if you’re under 16 but you can do REAL TIME if you’re over the age of 10. Get real @AlboMP.
0
2
5
RT @AndrewCMyers: Discovery is the reward. No one understands anything fully. Optimism pays off. Everyone has skills you lack. There are no….
0
16
0
Today's timeline is awash with "Please read my op-ed explaining why the Democrats would have won yesterday's US election, if only they had discarded the policies I disagrees with while promoting more of the policies that I prefer.".
0
0
2
“Optimism is a discipline. It's something that you put on. It's something that you seek and meet and wear and arm yourself with in order to endure, to persist, and to make life bearable and bearable for others.” Tim Winton, 2024.
0
0
1
A great and timely initiative. Props to all involved.
The review process is under threat and we must make it resilient against adversarial attacks. Proud to be part of @TheOfficialACM PROTECT where we'll look into securing academic reviewing
0
0
2
RT @tmiller_uq: I'm hiring for a postdoc position (women-only applicants) to work with the Ida Asadi Someh and Tapani Rinta-Kahila on desig….
0
6
0
I can attest last night’s 7pm show was indeed an excellent gig.
Adam and Wil The First Time kicked off last night with two awesome shows - thanks Melbourne!. Special guest was Rhys Muldoon. Who will it be in Sydney (Nov 9&10?). HINT; Wil may be sounding the Nerd Alert!. Some tix left @
0
0
1
To paraphrase Einstein, let me pose a provocation:. As far as mathematical specifications refer to real software they are not certain, and as far as they are certain they do not refer to real software.
I absolutely love the pdf example. It tells us that in the “real world” specs don’t really exist and we shouldn’t expect them to. If you are able to precisely specify correctness for your software, it means your software is weird (highly atypical, not pejorative).
1
0
1
This is true even for things like C compilers. CompCert is not even a counterexample because it is not a “C” compiler (not the way that users of Intel’s C compiler or some other arcane compiler understand what C means). see also
1
0
7
I absolutely love the pdf example. It tells us that in the “real world” specs don’t really exist and we shouldn’t expect them to. If you are able to precisely specify correctness for your software, it means your software is weird (highly atypical, not pejorative).
1
2
10
deep wisdom on why formal methods do and don’t “work” for practical projects and where, therefore, researchers should be focusing their attention. There is also an implicit argument here for gradual formal methods too. Fantastic presentation @miike.
I’ve done a lot of formal methods “technical sales” - trying to scope projects for Galois clients that solve their problem and fit into a budget. I gave a talk last week on some things that I learned along the way:
1
5
39
Kind of makes me wish I’d put a spoonerism into my DPhil thesis, though “Analysing the Pecurity Properties of Object-Capability Satterns” sounds almost astronomical.
Arrived at work this morning to find our department has a blue plaque! It commemorates the Rvd W A Spooner of "spoonerisms" fame, who lived in a terraced house that now forms part of our department in late victorian era. [1/3]
0
0
1
Predictions are a fool's game, so I'm happy to play: Tim Winton's "Juice" for the Booker 2025.
2
0
1
RT @eneelou: Cast your vote and help OWASP choose the next Top 10 risks for LLM security!. The voting process will also eliminate some entr….
0
3
0