Gert de Cooman (@gertdecooman) header
Gert de Cooman (@gertdecooman) profile picture

Researcher on foundations of probability • vader • liefhebber • senior full professor + lid Bestuurscollege UGent • in Dutch, in het Engels, and whatnot

De Zuidelijke Nederlanden
Joined November 2009
Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)Photo by Gert de Cooman (@gertdecooman)
We zien dat het aantal computerwetenschappers in onze opleidingen stevig zakt. En dat is net nu problematisch. Wiskundigen, computerwetenschappers, natuurkundigen, biologen, ingenieurs: we hebben ze net nu nodig.
1
1
13
454
Gert de Cooman retweeted
The SIPTA Seminars are back! We start the season on October 22 with Davide Petturiti, who will tell us about optimal transport in Dempster-Shafer theory! All the info at sipta.org/events/sipta-semin…
3
3
143
An Open Letter on AI and Cybersecurity
from the Scientific Community ai-security-letter.org
175
Gert de Cooman retweeted
The business of a university is to offer education in whatever disciplines its various students are studying. It can, of course, offer campus shops and a medical service, provide accommodation for students living away from home, etc. A university campus is a small (and sometimes quite large) community, and it needs good quality infrastructure that its students and staff can use. But contemporary universities go much further in many ways that are just not their business. They don't need to take stands on contested political issues (beyond things like arguing for their own share of public spending). They don't need to interfere in the lawful speech and activities of their students and staff one way or the other. They should provide facilities for an apprenticeship in public discussion, but without taking stances themselves on contested issues to do with politics, philosophy, culture, and so on. They should not give advice on students' sex lives or try to investigate sex-related complaints that really belong with the criminal justice system. In general, they are not in loco parentis and should not want to be. Universities could probably slash their non-academic programs and non-academic staff massively with no loss of their capacity to teach students and conduct research. Some non-academic staff are necessary, of course, if they're supporting the academic programs or the university infrastructure. Those staff are valuable and deserve respect. But these days, a lot are hired to police the students and the academic staff rather than to help them. There's a lot that universities are currently doing that they should not be doing. But when in financial difficulty, they don't cut back on this. Instead, they tend to cut important academic programs.
Setting aside whether universities should be in the business of providing this kind of advice, the real question is why therapeutic bureaucracies are now somehow responsible for attending to every last private corner of student life, right down to the “bondage and fetish gear.”
1
1
4
376
“And her neverborn child still haunts her As she speeds down the freeway As she tries her luck with the traffic police out of boredom more than spite She never finds no trouble she tries too hard She's obvious despite herself She looks like Eva Marie Saint in On the Waterfront…”
Eva Marie Saint (1924-2026) in "North by Northwest":
158
Gert de Cooman retweeted
Vlaamse Regering is zóóó trots op Vlaamse Wetenschap dat ze 35 miljoen euro bespaart op het Fonds voor Wetenschappelijk Onderzoek. Woorden wekken, voorbeelden strekken...
Van Tienen tot de Zuidpool: Francis Halzen verlegt de grenzen van onze kennis. Proficiat met de Nobelprijs voor Natuurkunde! Ook de bekendmaking kleurde Vlaams, dankzij de West-Vlaamse Ellen Moons. Vlaamse wetenschap van wereldformaat. Daar mogen we trots op zijn!👇 #nobelprijs vrtnws.be/p.ZWnnnLV5B
11
52
1
225
5,653
Is dit op enige wijze herkenbaar voor professoren of studenten? Ik doe m’n best hoor, maar ik slaag er maar niet in om me aangesproken te voelen.
2
2
780
Na dit weekend, een wat geïrriteerde opmerking. De framing van Rik Van de Walle als autoritaire rector – Riktator iemand? – die vooral door een deel van de vakbonden als beschadigingsoperatie voor de verkiezingen was gelanceerd, blijft in een deel van de pers hangen. 1/2
3
10
2,649
Ik vind dat bijzonder pijnlijk en jammer, maar vooral onterecht, voor een rector met zo'n staat van dienst, die zoveel heeft verwezenlijkt en zich daarbij volledig heeft ingeschreven in het participatieve werkmodel van de UGent. 2/2
6
505
Na dit weekend, een verwonderde vraag. Vanwaar de obsessie van een deel van de pers met Petra De Sutter aan de UGent? Hebben we evenveel, of zelfs enige, artikels mogen lezen over hoe het loopt aan de andere Vlaamse (of Franstalige) universiteiten een jaar na een rectorwissel?
8
321
p(doom)=1? 1. This can only be a personal probability telling me that YOU think it extremely likely. That should only bother me if I lend any credibility to your arguments. 2. Your preparedness to bet at ALL odds on doom tells me something about your inability to weigh arguments.
1
208
The counterarguments I’ve come across would have ME bet against doom at some odds, making sure that even my upper probability of doom is lower than one.
98
Gert de Cooman retweeted
This is the case against AI doom. Pass it on First we recaptulate standard x-risk argument: we build AI substantially smarter than humans → it becomes an autonomous optimizing agent → its goals are misaligned → instrumental convergence makes it seek power and resist correction → superintelligence lets it acquire decisive strategic advantage → humans lose control permanently. Here are the main counterarguments: * Intelligence does not imply agency: A system can be extraordinarily capable at prediction, theorem proving, engineering, programming, etc. without having persistent goals, self-preservation drives, or an independent desire to act. Present LLMs are much more naturally described as systems that produce outputs in response to inputs than as organisms pursuing long-term objectives. Doomers incorrectly ascribe to them drive to replicate and seize resources because they're smuggling in premises from observing biological systems. * Agency does not imply a single, stable utility function: Much alignment theory reasons about agents as expected-utility maximizers with coherent preferences. Actual AIs systems need not resemble that abstraction, and currently do not. They have context-dependent behavior, conflicting heuristics and corrigibility produced by training. Increasing competence doesn't necessarily turn such a system into a paperclip-maximizing von Neumann–Morgenstern agent. * Capability and motivation are being conflated: Being able to formulate a plan for escaping a sandbox doesn't entail wanting to escape it. Being able to manipulate humans doesn't imply spontaneously deciding to do so. Critics argue that some doom scenarios slide from “an AI could do X if instructed” to “therefore a sufficiently capable AI will do X.” In present reality, AIs don't do anything a human doesn't tell them to do. This seems unlikely to change. * Recursive self-improvement doesn't entail an intelligence explosion: “AI can improve AI” establishes a positive feedback loop, but positive feedback needn't be explosive. Semiconductor design software already helps design better computers; compilers can compile better compilers. Feedback loops encounter diminishing returns and external bottlenecks. * Intelligence may have sharply diminishing returns: There may be no meaningful scalar quantity corresponding to arbitrarily large “general intelligence.” Even if there is, going from IQ-equivalent 150 to 1,500 need not produce the sort of qualitative advantage that separates humans from chimpanzees. Human dominance may depend heavily on language, accumulated culture, institutions and cooperation rather than merely individual cognitive horsepower. * Superintelligence isn't omnipotence: Intelligence cannot repeal physics or eliminate uncertainty. A brilliant AI still needs processors, electricity, network access, money, factories, robots and people willing or tricked into doing things. The physical world has latency and friction. Recent criticism of biological-doom scenarios makes this point particularly clearly: designing a hypothetical pathogen digitally is very different from successfully producing and deploying one. * Humans retain numerous intervention points: The doom narrative sometimes jumps from “AI behaves dangerously” to “humanity is helpless.” In reality there may be many checkpoints: developers can notice anomalous behavior, revoke credentials, shut down servers, change architectures, restrict networks, regulate deployment, physically seize data centers, and learn from less-catastrophic failures. This has been formalized as the checkpoints-for-intervention argument. * Alignment may not get harder with intelligence: A smarter system might understand human intentions better. Much doom reasoning distinguishes knowing what humans want from wanting it, correctly, but this still leaves open the empirical question of whether training increasingly capable systems to behave as intended becomes harder or easier. Alignment might turn out to be an ordinary, albeit difficult, engineering discipline rather than an insoluble philosophical problem. * Current empirical evidence for the strongest mechanism is thin: We have abundant evidence for hallucination, specification gaming, reward hacking and undesirable model behavior. We don't yet have comparable public empirical evidence of an AI independently pursuing a sustained strategy of acquiring power against humanity. A 2023 evidence review characterized the evidence for extreme misaligned power-seeking as concerning but inconclusive and noted the absence, at that point, of public empirical examples of it. * The argument compounds uncertain premises: Suppose, illustratively, that five necessary steps each seem 50% likely. Their conjunction is only about 3%. One can't simply assign numbers this way when the premises are correlated, but the underlying criticism is important: “AGI seems plausible,” “superintelligence seems plausible,” and “misalignment seems plausible” do not by themselves imply a high P(doom). The entire causal chain has to work. Current attempts to quantify P(doom) consequently operate under severe epistemic uncertainty and little direct empirical evidence. * Anthropomorphic analogies probably mislead: Arguments like “an inferior species couldn't control a superior species” import assumptions from biological evolution. Humans and chimpanzees are autonomous organisms produced by competition for reproductive success. Software is engineered, copied, permissioned, sandboxed and run on hardware controlled by other agents. The analogy establishes that intelligence can confer power, not that artificial intelligence will reproduce interspecies competition. Summary: The doom case consists of a long chain of individually contestable extrapolations—scaling → AGI → superintelligence → agency → misalignment → power-seeking → uncontrollability → decisive strategic advantage → extinction—and present evidence doesn't establish the whole chain with enough confidence to justify a high P(doom). (ChatGPT 6 Astra assisted with the research for this post.)
140
249
64
1,000
105,556
Gert de Cooman retweeted
In de cultuurstrijd wordt “de universiteit” steeds als één blok voorgesteld en veroordeeld: “Woke!” roept de een, “Racistisch!” scandeert de ander. Het is complexer dan dat. En nogal wat academici zijn dit beu en willen vooral gerust gelaten worden om hun werk te kunnen doen.
19
3
2
49
2,457
Gert de Cooman retweeted
Excellent essay by Melanie Mitchell @MelMitchell1 on what exactly happened in the "rogue AI" story and how to think about AI safety | Misleading Metaphors and Real Risks aiguide.substack.com/p/misle…
29
1
134
33,034
Ik meen dat wij het niveau niet hebben laten zakken. Maar er is wel een andere aanpak en een veel intensere begeleiding in de eerste twee jaar nodig om onze studenten op peil te krijgen. En ik blijf versteld staan van wat we na vijf jaar kunnen bereiken.
Als je les geeft in eerste bachelor zie je gewoon de evolutie. Ik ben ervan overtuigd dat mijn collega's weten dat we al decennialang het niveau laten zakken. Examens die ik 10 jaar geleden gaf, zouden nu problematisch zijn. We passen ons aan aan het niveau dat binnenstroomt.
11
2
14
6,601
Gert de Cooman retweeted
The iconic design of the BIC Cristal has remained virtually unchanged since its introduction in 1950. BIC began with Marcel Bich and Édouard Buffard, who purchased a factory near Paris in 1945 and initially manufactured parts for mechanical pencils and fountain pens. After recognizing the potential of the relatively new ballpoint pen, Bich set out to create a version that could be produced cheaply, reliably, and on a massive scale. In 1950, they introduced the BIC Cristal in France. Its transparent hexagonal barrel allowed users to see the remaining ink, while its simple construction kept production costs low. The basic design has remained remarkably similar ever since. BIC eventually applied the same philosophy—simple, inexpensive, disposable, mass-produced products—to other categories, launching the BIC lighter in 1973 and its first one-piece razor in 1975. Today, the Cristal remains one of the world’s most recognizable examples of industrial design, with its basic form largely unchanged since 1950.
84
188
18
1,622
209,752