Semantic Computing withย IEML
Follow
Click to Follow plevy
Pierre Levy 
@plevy
Pierre Levy ๐ปโs Tweets
๐ฒ
Quote Tweet
On the highway towards Human-Level AI, Large Language Model is an off-ramp.
Show this thread
1
How do spatial collectives harness physics to produce (intelligent) adaptive output?
Shishkov+ review physical principles influencing superorganism emergence, incl insights from adaptive (non-living) matter+invertebrate soft matter systems. 6/9
journals.sagepub.com/doi/full/10.11
1
23
68
Show this thread
How do science knowledge growth processes relate to collective innovation?
et al find concept networks grow by creating+filling gaps rather than expanding knowledge outwards+ r awarded for discoveries in these spaces.
journals.sagepub.com/doi/full/10.11
1
4
17
Show this thread
Diversity can contribute to collective intelligence but can also lead to polarization + impede innovation diffusion. Recent papers published in COLLECTIVE INTELLIGENCE () explore how beliefs spread, influence of outgroups + role of hubs 1/9.
journals.sagepub.com/articles/COL
1
25
89
Show this thread
ยซย The Symbolic Speciesย ยป, by T. W. Deacon, even if 25 years old, is excellent. A classic in evolutionary neuro-cognitive sciences.
4
24
Ya pueden verlo completo. Les sugiero lean la presentaciรณn del nรบmero especial de y de un servidor. twitter.com/REDDUSC/status
This Tweet is unavailable.
3
3
"An imaginative way to incorporate AI-generated text into teaching could be for the teacher to employ Transformer AI to generate a set of alternative essays on a topic, then ask students to critique these and write their own better versions."
Quote Tweet
Replying to @KTDigital
Well, you might try my open access opinion piece in Journal of AIED. It begins with a brief explanation, then an example of a "student essay" written with GPT-3, then some creative ways to use AI language generators for education.link.springer.com/article/10.100
1
6
14
ยซย Language models are not knowledge basesย ยป Let that sink in.
Quote Tweet
Heard a group of students in the sauna today. All are using #ChatGPT for assignments but worried about staff catching them with detector tools. Unis urgently need policy for ethical use of AI tools, with clear guidelines for students. Here are draft guidelines, courtesy of GPT.
Show this thread
4
6
21
"trouver le passage entre un rรฉalisme dogmatique, pour qui un รชtre objectif serait indรฉpendant de ceux qui lโapprรฉhendent, et un relativisme irresponsable selon lequel il nโy aurait de rรฉalitรฉ que pour un observateur ou un penseurโฆ "
qlq chose entre 1 et 0 (sourire)ยฒ
Quote Tweet
"Une espรจce de primates supporte un รฉcosystรจme symbolique : voilร lโhumanitรฉ." Un texte de 2019 in @SensPublic
sens-public.org/articles/1396/
2
1
"Une espรจce de primates supporte un รฉcosystรจme symbolique : voilร lโhumanitรฉ." Un texte de 2019 in
2
7
#culture and #collectiveintelligence in societies of animals, humans and robots | Philosophical Transactions of the Royal Society : Biological Sciences
1
4
13
Les IA gรฉnรฉratives ont crรฉรฉ une situation irrรฉversible. Mais attention ร la censure, aux biais algorithmique et ร la normalisation des images (et des textes).
Quote Tweet
On a demandรฉ ร รtienne Mineur, designer rรฉputรฉ pour รชtre l'un des premiers ร travailler avec les intelligences artificielles, ce qu'il pense des bouleversements crรฉรฉs par Dall-E et Midjourney.
ladn.eu/mondes-creatif via @LADN_EU
2
5
The reading for my next book has begun!
2
8
40
METAVERSE AND EDUCATION
It is ready to be published on January 31 at revistas.um.es/red/issue/view with DOI doi.org/10.6018/red
Guest editor
Presentation red.hypotheses.org/2271 and red.hypotheses.org/2330
#MetaverseEducation
8
11
Aprรจs avoir fait IEML, je rรฉflรฉchis ร mon prochain livre. Un titre possible ยซย Lโintelligence humaineย ยป Sous titre: ยซย Pensรฉe conceptuelle et production de sensย ยป.
4
7
30
The AI alignment problem: ยซย This is essentially the old story of the genie in the lamp, or the sorcerer's apprentice, or King Midas: you get exactly what you ask for, not what you want.โ
2
5
23
Ya estรก casi listo el nรบmero especial de RED sobre #metaverso_educaciรณn. Editor invitado . Con importantes conclusiones.
El 31 de enero. Tercer nรบmero que publica RED en un mes.
DOI: doi.org/10.6018/red, activo el 31 de enero
7
9
ยซย LLMs show impressive (although imperfect) performance on tasks requiring formal linguistic competence, but fail on many tests requiring functional competence. For modeling thought, models would need to incorporate multiple non-language-specific cognitive capacities.ย ยป ๐
Quote Tweet
Three years in the making - our big review/position piece on the capabilities of large language models (LLMs) from the cognitive science perspective.
Thread below! 1/
arxiv.org/abs/2301.06627
Show this thread
4
11
Sorry ยซย If I *were* still prof.ย ยป
9
Show this thread
If I was still prof. I would make the use of ChatGPT mandatory for writing essays. The students with the best prompts, dialogue with the system, choice of text and verification of content would have the best marks!
7
28
71
Show this thread
ยซย The hallucinogenic and fabulist quality of dreams means they are different from the โtraining setโ of the animal (i.e., their daily experiences). It is the very strangeness of dreams in their divergence from waking experience that gives them their biological function.ยป
Quote Tweet
Erik Hoel proposes something Iโve suspected all along: dreams are data augmentation!
In dreams, your brain mashes up, permutes, and adds random variation to your memories, so you donโt โoverfitโ excessively rigid models of your experience.
sciencedirect.com/science/articl
Show this thread
3
13
Last week, & hosted #AGIDebate featuring AI powerhouses for 4hrs! We learned a TON + are sharing our notes w/ y'all!
Included:
โฆ Profiles & timestamps
โฆ Notes & short summaries for all speakers
โฆ Comments enabled!
โค๏ธ๐ค๐
2
21
29
Show this thread
New Nature paper provides evidence that science has become less innovative since the 1950s. The authors suggest reversing the trend by:
1. reading widely,
2. focusing less on quantity of papers, & more on research quality,
3. taking year-long sabbaticals.
nature.com/articles/s4158
373
4,305
14K
Show this thread
Just wait for the dance-code performance!
Quote Tweet
Light Balance Kids brought the light to #AGTAllStars and @howiemandel gave them his #GoldenBuzzer! 
7:42
218.7K views
1
2
7
Nessuno conosce tutto,
tutti conoscono qualcosa,
tutta la conoscenza
abita nell'umanitร
Pierre Lรฉvy
#AspettandoCapodanno #SalaLettura
1
20
35
Show this thread
ยซย Algorithms do not simply discern what we want and serve it to us; they train us to want what they can serve us. Successful platforms do not just discover what consumers want โ they produce the consumers and the forms of consumer desire that they need.ย ยป
Quote Tweet
This article by โฆ@wrbuclaโฉ is probably not a self-evident read for many, but it really is a must-read, punching through narrative and mythos alike.
โDigital abundance is a mixed blessing: exhilarating, yet flattening and homogenizing.โ
Atomising. noemamag.com/hyperconnected
15
38
Chat-GPT renvoie lโรฉcho de la caverne (de Platon).
Quote Tweet
[New Blog Post] GPT-3 : c'est toi le Chat. affordance.framasoft.org/2023/01/gpt-3-
3
4
4
Cet aprรจs-midi ๐จ๐ฆ
2
2
36
Happy New Year๐ะฉะฐัะปะธะฒะพะณะพ ะะพะฒะพะณะพ ัะพะบั ๐ุณุงู ูู ู
ุจุงุฑฺฉ ๐ฅนBonne et heureuse annรฉe!๐ ุณูู ุฌุฏูุฏู ุณุนูุฏู ฮฯ
ฯฯ
ฯฮนฯฮผฮญฮฝฮฟ ฯฮฟ ฮฝฮญฮฟ ฮญฯฮฟฯ โค๏ธโ๐ฅBon any nou ืฉื ื ืืืื Jabulela unyaka omusha เฆถเงเฆญ เฆจเฆฌ เฆฌเฆฐเงเฆท ะจะธะฝั ะพะฝั ะผัะฝะด ั
าฏัะณัะต ใใใพใใฆใใใงใจใ ์ํด ๋ณต ๋ง์ด ๋ฐ์ผ์ธ์ โจแแธแแแถแโแแแแถแโแแแแธ
2
3
24
Expert systems, traditionally classified under the rubric of "artificial intelligence", should be considered as techniques of communication and rapid mobilization of practical skills in organizations rather than as doubles of human experts. Images dreamstudio.ai
read image description
ALT
read image description
ALT
read image description
ALT
read image description
ALT
1
2
5
Show this thread
TECNOLOGIA
Inteligรชncia Artificial produz "monstruosidades รฉticas", diz Pierre Lรฉvy
Filรณsofo avalia que estamos longe de uma tecnologia capaz de produzir verdadeiro conhecimento ou entendimento do mundo
bit.ly/3VrY9no
20
95
Hilma af Klint
Sweden, 1862-1944
(The woman who discovered abstract art before Kandinsky).
3
22
91
Because statistical learning is not enough to reach a real practical knowledge of the world. We need also better knowledge representation, reference, reasonning and effective language understanding (see my pinned post). Statistical learning is only one part of the solution. IMHO.
1
2
13
Show this thread
was thinking about our semiotic trails and the way we make sense of them with a view to collective intelligence long before I realized the connection between our private and public lives from the perspective of data.
Enjoy the Dialogue (real one)
6
7