Sign in

James McDermott

@jmmcd.bsky.social
212 followers 350 following 198 posts

Senior Lecturer in Computer Science, University of Galway, Ireland. I'd rather write programs that write programs than write programs.

PostsRepliesMedia
James McDermott @jmmcd.bsky.social · 13/02/2026
What I also find weird is that according to canon, they are just ordinary robots. Transforming to cars and planes is just something they did when they had to visit Earth for some reason.
030
James McDermott @jmmcd.bsky.social · 06/02/2026
Aha, thank you, yes I have heard it is great. Moving Pantheon to the top of my list now that Pluribus is over until the distant future (late 2027, likely post-singularity for all I know)
010
Reposted by James McDermott
RTÉ News @news.rte.ie · 03/02/2026
French police have searched the offices of Elon Musk's social ⁠media network X, and prosecutors ordered the tech billionaire to face questions in April related to a widening investigation into the platform, the Paris prosecutor's office said.
rte.ie
French cybercrime authorities search X offices
French police have searched the offices of Elon Musk's social ⁠media network X, and prosecutors ordered the tech billionaire to face questions in April related to a widening investigation into the pla...
14115
James McDermott @jmmcd.bsky.social · 04/02/2026
A (very good) drama about The Singularity sounds like it matches my preferences - reco please?
100
James McDermott @jmmcd.bsky.social · 19/01/2026
What I don't get is I thought Trump had already used and discarded Musk, had fully fallen-out with him. Why still use Grok? I also admit I don't understand whether Trump's Greenland thing is further proof, or is disproof, of him working for Russia. He is a first-class chaos monkey, unreadable.
110
James McDermott @jmmcd.bsky.social · 15/01/2026
This is a common pattern, where leftists pretend that people who become hateful were never talented in the first place. Another example: Graham Linehan. I say this as a concerned leftist - concerned about people's honesty. These people should re-read Aesop's Sour Grapes.
000
James McDermott @jmmcd.bsky.social · 15/01/2026
It's very interesting in the age of Claude Cowork to look back a year or two to see "sensible, anti-hype" articles which were wrong at the time but are now demonstrated to be wrong. A lesson for all people who think anti-hype is automatically sensible. Example: cacm.acm.org/blogcacm/how...
cacm.acm.org
000
Reposted by James McDermott
Dr Abeba Birhane @abeba.blacksky.app · 15/01/2026
are you displeased with today’s AI safety evaluation landscape and curious about what greater conceptual clarity, methodological soundness, and rigour in AI evaluation could look like? if so, consider coming to Dublin to pursue a PhD with me apply here: aial.ie/hiring/phd-a... pls repost
About the PhD: 
Audits and evaluation of AI systems — and the broader context that AI systems operate in — have become central to conceptualising, quantifying, measuring and understanding the operations, failures, limitations, underlying assumptions, and downstream societal implications of AI systems. Existing AI audit and evaluation efforts are fractured, done in a siloed and ad-hoc manner, and with little deliberation and reflection around conceptual rigour and methodological validity.

This PhD is for a candidate that is passionate about exploring what a conceptually cogent, methodologically sound, and well-founded AI evaluation and safety research might look like. This requires grappling with questions such as:

    What does it mean to represent “ground truth” in proxies, synthetic data, or computational simulation?
    How do we reliably measure abstract and complex phenomena?
    What are the epistemological or methodological implications of quantification and measurement approaches we choose to employ? Particularly, what underlying presuppositions, values, or perspectives do they entail?
    How do we ensure the lived experiences of impacted communities play a critical role in the development and justification of measurement metrics and proxies?
    Through exploration of these questions, the candidate is expected to engage with core concepts in the philosophy of science, history of science, Black feminist epistemologies, and similar schools of thought to develop an in-depth understanding of existing practices with the aim of applying it to advance shared standards and best practice in AI evaluation.

The candidate is expected to integrate empirical (for example, through analysis or evaluation of existing benchmarks) or practical (for example, by executing evaluation of AI systems) components into the overall work.
5191139
Reposted by James McDermott
Bill Tozier @vaguery.bsky.social · 13/01/2026
Occasional reminder that I sell books and similar stuff on eBay. Link to my store is below. Just reached 4100+ items for the first time. So there are categories and internal search in the storefront, which might help a bit. www.ebay.com/sch/i.html?_...
ebay.com
Items for sale by cornersbumpedbooksandantiques | eBay
Shop eBay for great deals from cornersbumpedbooksandantiques!
163
James McDermott @jmmcd.bsky.social · 13/01/2026
I do sometimes use a 3B model locally via Ollama. It's already magical at that small scale. I'm confident that small improvements in training will give us 3B coding models good enough to go completely off-grid, for example.
000
James McDermott @jmmcd.bsky.social · 13/01/2026
Speech versus text is a big deal, but we already have more than enough bandwidth for that (actually, we had that a century ago).
010
James McDermott @jmmcd.bsky.social · 13/01/2026
From my perspective, avatars would not add anything. LLMs at current levels of capability and current internet speeds/bandwidths are already transformational and there is a huge amount to be done with current models - no way that a lack of avatars makes LLMs flop.
300
James McDermott @jmmcd.bsky.social · 12/01/2026
Genetic Algorithms General Evolutionary Computation and Hybrids Genetic Programming Learning for Evolutionary Computation Neuroevolution Real World Applications Swarm Intelligence Theory
000
James McDermott @jmmcd.bsky.social · 12/01/2026
Tracks including: Benchmarking, Benchmarks, Software, and Reproducibility Complex Systems Evolutionary Combinatorial Optimization and Metaheuristics Evolutionary Machine Learning Evolutionary Multiobjective Optimization Evolutionary Numerical Optimization
100
James McDermott @jmmcd.bsky.social · 12/01/2026
GECCO call for papers - abstract deadline Jan 21st! gecco-2026.sigevo.org/Tracks @skapedin.bsky.social @folivetti.bsky.social
100
James McDermott @jmmcd.bsky.social · 12/01/2026
That is provocative - why do you think it matters?
100
James McDermott @jmmcd.bsky.social · 19/12/2025
The HUMIES are the annual competition at #GECCO for human-competitive results created by evolutionary computation methods. ***Ten thousand dollars in prizes for human-competitive results*** Submit by Friday May 29, 2026 See www.human-competitive.org for more
A cartoon of a strand of DNA, arm-wrestling a scientist
000
Reposted by James McDermott
Gavan Reilly @gavreilly.com · 05/10/2025
Wrote a Substack on this #aras25 poll: Catherine Connolly is surviving her scrapes - yet Heather Humphreys is within the margin of error for winning open.substack.com/pub/gavreill...
open.substack.com
Today's opinion poll merely shows how close this Áras race actually is
Giving the Google Sheets treatment to the Sunday Times/Ireland Thinks poll
3237
James McDermott @jmmcd.bsky.social · 04/10/2025
You should try reading comments on Imgur! Some of them want to abuse you All of them will disabuse you (of that notion)
010
Reposted by James McDermott
GPEM journal @gpem.bsky.social · 04/10/2025
New special issue of GPEM on Evolutionary Computation in Art, Music and Design! Edited by Penousal Machado and Juan Romero link.springer.com/article/10.1...
link.springer.com
Editorial Introduction to the Special Issue on Evolutionary Computation in Art, Music and Design - Genetic Programming and Evolvable Machines
Genetic Programming and Evolvable Machines -
111
Reposted by James McDermott
GPEM journal @gpem.bsky.social · 21/08/2025
New book review, freely available in GPEM: “Reversible world of cellular automata” by Kenichi Morita, reviewed by Tomas Rokicki link.springer.com/article/10.1...
link.springer.com
Kenichi Morita: Reversible world of cellular automata - Genetic Programming and Evolvable Machines
Genetic Programming and Evolvable Machines -
074
James McDermott @jmmcd.bsky.social · 06/07/2025
[5] Synth line uses Flow (Sean Luke) github.com/eclab/flow [6] Vocoder is Logic Pro EVOC [7] Lyrics where the narrator dreams about a heist going wrong: Kate Bush, "There Goes A Tenner"
010
James McDermott @jmmcd.bsky.social · 06/07/2025
Bibliography: [1] Intro texture and fade-up: Metallica, "Orion" [2] Drum hitch: Stone Roses, "I Wanna Be Adored" and Cranberries, "Dreams" [3] Kick/snare inversion: The Cars, "Just What I Needed" [4] Chord sequence played on Tonnetz (Euler, 1739) implementation at thetonnetz.com
thetonnetz.com
The Tonnetz
100
James McDermott @jmmcd.bsky.social · 06/07/2025
youtu.be/qhJrs85be64
youtu.be
Bad Dreams / I Wanna Be Abhorred
YouTube video by James McDermott
100
James McDermott @jmmcd.bsky.social · 06/07/2025
Bono is an easy target and whatever about the UK, I bet a lot of Irish people would line up to say they agree with this article. I think he's brilliant.
021
Reposted by James McDermott
GPEM journal @gpem.bsky.social · 02/07/2025
GPEM journal has a new special issue on "twenty-five years of grammatical evolution"! Edited and with an introduction by Mahdinejad, Murphy and Ryan. Special issue: link.springer.com/collections/... Introduction: link.springer.com/article/10.1...
link.springer.com
Special Issue on Twenty-Five Years of Grammatical Evolution
By invitation only- GECCO conference ("GEWS2023 — Grammatical Evolution Workshop)
111
James McDermott @jmmcd.bsky.social · 02/07/2025
Scalar mappings seem wrong. I think length, volume, mass, and time all map differently.
000
James McDermott @jmmcd.bsky.social · 30/06/2025
Maybe itinerant fish intermittently school in the intertidal zone underneath the waves of engagement
010
James McDermott @jmmcd.bsky.social · 30/06/2025
Itinerant -> intermittent There you go, the kind of engagement we all long for
120
James McDermott @jmmcd.bsky.social · 19/03/2025
Our gas cooker is an annoying sigmoid. The gas tap is either on with minimal flow, or on with full-flow, with only a tiny range - about 5 degrees of the controller's turn - giving useful intermediate values.
000
James McDermott @jmmcd.bsky.social · 18/03/2025
Next up: Massachusetts
010
James McDermott @jmmcd.bsky.social · 18/03/2025
Ah ok. Obvious in retrospect. Yes, I agree. Intellectual tasks turned out to be "easy" because iterating in software is so much faster. Kind of sad, isn't it? I wanted AI to do the housework.
130
James McDermott @jmmcd.bsky.social · 18/03/2025
Thought experiment: compare (1) an end-to-end LLM reading a problem statement and providing an answer, versus (2) translate the problem to PKN, then run formal reasoning over the PKN, then translate the result back to English. My intuition says a good LLM will win this.
000
James McDermott @jmmcd.bsky.social · 18/03/2025
Maybe the PKN becomes a symbolic layer, a bit like the system 2. But LLMs already have words (system 2) on top of their statistical/lower-level vectors (system 1). I'm not convinced that reasoning in PKN can do better than reasoning in LLMs.
100
James McDermott @jmmcd.bsky.social · 18/03/2025
In a way, it's weird that we are still trying initiatives like this. We've already shown that LLMs are sufficient for a lot of reasoning, and moreover, anyway, are necessary (translating PKN to/from English is a job for LLMs, not for humans).
100
James McDermott @jmmcd.bsky.social · 18/03/2025
Plausible Knowledge Notation: "a notation and model for imperfect knowledge, such as everyday knowledge that is uncertain, context sensitive, imprecise, incomplete, inconsistent and likely to change" w3c.github.io/cogai/pkn.html
w3c.github.io
Plausible Knowledge Notation (PKN)
100
James McDermott @jmmcd.bsky.social · 18/03/2025
What does "durable" mean here?
100
James McDermott @jmmcd.bsky.social · 10/02/2025
Congratulations!
010
Reposted by James McDermott
Dr Robert Bohan @robertbohan.bsky.social · 10/02/2025
Ireland! We need just 1,483 ppl who are Irish nationals to sign this EU petition to initiate a ban on LGBTQ conversion therapy* across the EU. The UN considers such ‘therapy’ to be mental torture. #SpeirGorm Please sign & share with all EU citizens that you know: eci.ec.europa.eu/043/public/#...
eci.ec.europa.eu
European Citizens' Initiative
Give your support !
18360211
James McDermott @jmmcd.bsky.social · 13/01/2025
Students who can't really write in English but use LLMs to write their emails to me - should I treat them as having communicative intent? Do they intend to say the things they say, or do they intend some underlying meaning which I should charitably guess at? Or is it just empty syntax?
000
James McDermott @jmmcd.bsky.social · 24/12/2024
I just saw Thomas Riker try to bluff his doppelgänger Will in 5-card draw, and fail dismally (TNG season 6). Not entertaining except as sadism, 2/10.
010
James McDermott @jmmcd.bsky.social · 22/12/2024
Lots of people do dismiss concerns, and in fact two main sets of concerned people dismiss each other's concerns. The point here is that Ethan does not. Your comment is toxic and deserves a block.
010
James McDermott @jmmcd.bsky.social · 21/12/2024
There are some arguments either way on the bottom line, but these Nazi IQ tests surely constitute zero evidence either way.
010
James McDermott @jmmcd.bsky.social · 21/12/2024
I think you can be heterodox / anti-woke (in a specific definition of that contested term) without going anywhere near the space lasers or the right.
220
James McDermott @jmmcd.bsky.social · 21/12/2024
Any task which is so easy for humans and so hard for computers is necessary, but not sufficient, for AGI. It was originally just called ARC (5 years ago).
010
James McDermott @jmmcd.bsky.social · 21/12/2024
Do you mean "works" as in "produces a smaller model of same performance" (seems easy to accept) or ".. of better performance"?
100
Reposted by James McDermott
Iris van Rooij 💭 @irisvanrooij.bsky.social · 18/12/2024
Dear international colleagues, repression in the Netherlands is worsening. We can really use your support. Please join me in signing this open letter demanding accountability from our Executive Board for the wrongful arrest of six students and staff at my university docs.google.com/forms/d/e/1F...
docs.google.com
Hold Radboud University accountable for the wrongful arrest of staff and students
To the Radboud College van Bestuur, On Monday December 16, six staff and students from the Radboud University were wrongfully accused of property damage and held overnight in jail on the grounds of s...
25823
James McDermott @jmmcd.bsky.social · 21/12/2024
Signed. But note the video seems to be wrong in its two claims about law. 1. Arrest is not prosecution. 2. "Promptly informed" does not mean in advance, it allows for that to happen soon after, eg at the station. I wonder what the crime reports looked like.
000
James McDermott @jmmcd.bsky.social · 17/12/2024
That link is 404 now
200
Reposted by James McDermott
Jacob @tullamore-jew.bsky.social · 16/12/2024
I could write an essay on this image Satire of antisemitic cartoons made to mock antisemitism repurposed by an Israeli to depict the Irish as the shadowy backers of their non-white enemies, thus bringing the caricature's functionality back to its original intent
Caricature of an Irish man done in the style of old antisemitic cartoons. He has bright ginger hair, is haunched over with clenched hands together. He is holding a Palestinian flag and wearing a Hamas headband.
1613427