Inverse Laws of Robotics:
* Humans must not anthropomorphise AI systems.
* Humans must not blindly trust the output of AI systems.
* Humans must remain fully responsible and accountable for consequences arising from the use of AI systems.
Remote
Mathieu Jacomy
@jacomyma@mas.to
mastodon 4.7.2Digital tools + social sciences + design at the Tantlab in Copenhagen. Co-creator of Gephi & Hyphe.
1018 Followers
392 Following
31 Posts
Joined October 04, 2022
Loved debating whether ecologists should fly across the world to the big conferences where the powerful meet?
You'll love: Can we use LLMs to critique LLMs at scale?
Matt Artz' "gephi-ai" allows AI agents to remote-control Gephi and perform an analysis.
I just tested it and I was impressed. Claude Code actually looked at rendered images to iterate over the settings. It also identified something many students miss about this network: its time dimension. And it aligned the map vertically as I had asked.
The install was a bit tedious to install, but it's just an alpha version for now.
The repo: https://github.com/MattArtzAnthro/gephi-ai
My install and test walkthrough: https://gist.github.com/jacomyma/754e8995592025b4fa897dd75617421b
Apparently a Dane born in Copenhagen is a person who, once drunk, realizes they can speak Danish.
Any technology that promises you more independence is in fact just offering to trade a visible dependence for an invisible one.
(STS shower thought)
RE: @Yomguithereal@mastodon.social
xan is a command line tool that can be used to process CSV files directly from the shell.
Independent execution of computations underlying research articles.
https://codecheck.org.uk/
I just finished rendering the "grounded AI map", a giant map of AI and algorithms in science that we are going to print as a 100m² vinyl floor mat that one can walk on, and exhibit in different places (stay tuned).
Just showing a few pics because it's gorgeous.
Each dot is a paper. The color is the year (red 1980 to blue and purple 2020+).
Clusters are labelled + a code number allows getting more info.
"AI is an incredible force multiplier for implementation, but it’s a dangerous substitute for design."
Excellent, honest report on carrying out a software engineering project using agentic AI.
Replying to
Link from Hacker News. Discussion there FYI: https://news.ycombinator.com/item?id=48307887
Replying to
I struggle to have my students articulate their findings without telling the story of their journey. That's because a reader does not want to give her attention to the "how we got there" until you've convinced her that "where we got" was worth it.
Replying to
But narrating the journey has a perverse effect on some of my students: it lets them believe that their analysis' legitimacy derives from the amount of effort they have engaged in their work.
Replying to
@ewout@mastodon.social Absolutely. To some extent, the level of detail provided (version of ChatGPT) suggests that it has some relevance; but if it explains features of the writing, doesn't it defeat the claim that the authors take full responsibility?
The entire statements sounds like an apophasis to me.
Note: I will link the Wikipedia page for Apophasis because I did not know the word in English (we call it "prétérition" in French) so no shame here. The concept is worth knowing.
https://en.wikipedia.org/wiki/Apophasis
There Is No ‘Hard Problem Of Consciousness’
https://www.noemamag.com/there-is-no-hard-problem-of-consciousness/
I will be covering the Open Research Devroom at the #FOSDEM all day, just follow the devroom's account!
I'm in a conference about networks and as often, some speaker says something like "I did not use Gephi for this, sorry".
I realize that my de-facto joke in this context has become "Now you need to make a donation to Gephi to absolve your soul."
That's so great to have the donate system functional again.
Replying to
BTW here are preview screenshots of our work in progress of integrating a new icon set, including a number of custom-made ones, into #gephi .
🦶 Step on AI! Trample our giant visualization!
The Grounded AI Map is a 100m² floor mat representing 2M scientific papers involving AI and algorithms. The Danish Technical Museum will be exhibiting it until June 2, 2025, in dialogue with objects selected from its collections.
The museum is in Helsingør (think Hamlet), north of Copenhagen.
Check it out:
https://youtu.be/n4XSpCz3iJg
Hey les Parisiens, I will be in Paris throughout February.
I'm visiting the Institute for Advanced Study to work on AI-augmented mapping and notably within RAG systems for computational SSH, along with @AndersKMunk@mas.to .
https://www.linkedin.com/feed/update/urn:li:activity:7420054429581881346/
Let's get in touch!
Tomorrow I will be at the Computational Humanities Conference in Aarhus to present a short paper as a poster, where we (Matilde Ficozzi, @AndersKMunk@mas.to & I) explain why some bridges in a network map can actually be clusters.
Sub-optimal Recall in Visual Cluster Retrieval: When Clusters Look Like Bridges
https://2024.computational-humanities-research.org/papers/paper5/
The Techno-Anthropology Lab, where I work, has made a video that is far too fun to stay buried in the sad shell of a website that the university has forced upon us, so I share it here. Enjoy!
🛠️ They make tools for the social sciences and humanities. Who are they? What do they want? Why don't they make better, different tools?
On October 12 from 16:00 to 18:00 in Copenhagen, MASSHINE hosts a panel to give a voice to the SSH tool makers. Book your agenda and come discuss with us! Just register in the link below.
https://www.en.ssh.aau.dk/masshine-talk-computational-tool-making-for-the-social-sciences-humanities-e85671
Replying to
But making an argument is too difficult for some of my students, and they fall back to narrating their struggle instead. It pushes back onto me the responsibility to delineate the argument. I will do it as a teacher, but not as a reader. It's the writer's job to make a point.
University logic
"Sacit" vs. "Saphor" data viz framework.
Sacit: simple, authoritative, complete, impervious to the viewer, and transparent.
Saphor: situated, authored, partial, historical, observer-dependent, and rhetorical.
I'm just offloading a discarded fragment I had drafted during my writing process. The second series of qualifiers is from Johanna Drucker.
Working too much with text embeddings.
Vibe coding is maieutic.