Institut Jean Nicod

Accueil du site > Séminaires & Colloques > Colloquium de l’IJN > Colloquium



Colloquium

 

2019-2020

 


 

Lisa Miracchi (University of Pennsylvania)

Vendredi 13 décembre 2019 de 11h30 à 13h

Institut Jean-Nicod, Pavillon Jardin, ENS, 29, rue d’Ulm 75005. Salle de réunion, RDC

"Generating Intelligence : How AI Can Make the Leap from Artifacts to Agnets"

Abstract

While current AI technology has provided us with useful tools for accomplishing our aims, we have yet to build anything that plausibly has, and can robustly and autonomously execute, its own aims. I argue for a systematic approach to this project that respects the ways in which genuine agents behave differently form other kinds of systems. Not only do they exhibit different kinds of behavioral patterns, the variables that are appropriate for describing and systematizing their behavior are importantly different, and typically involve interactions at larger spatiotemporal scales. Any account of the psychology of artificial intelligenve - what it is for an artificial agent to have beliefs and desires - must therefore be updated to account for their role in behavior described with these different variables. I argue that this provides us with a more accurate and plausible account of the kinds of mental states and processes that we mst implement in artificial systems if they are to exhibit the kinds of robustness, flexibility, and autonomy that even relatively simple animals regularly demonstrate. I then show how the Generative Methodology I have developed elswhere can be employed to structure research into building such systems.

 

 


 


 

Barbara Tversky (Columbia)

Vendredi 20 décembre 2019 de 11h30 à 13h

Institut Jean-Nicod, Pavillon Jardin, ENS, 29, rue d’Ulm 75005. Salle de réunion, RDC

"Putting Messy Thought in the World : Sketches and Perspective"

Abstract

Designers begin with messy sketches, scientists with messy data. Scrutinizing the messiness can lead to reconfiguration and new ideas, a virtuous cycle that can be encouraged by perspective-taking.

 


 

Markus Werning (Ruhr University Bochum)

 

"Neither Preservation, nor Imagination : Episodic Memory as a ’Prediction of the Past’ from Minimal Traces"

Vendredi 20 mars 2020 de 11h30 à 13h

Institut Jean-Nicod, Pavillon Jardin, ENS, 29, rue d’Ulm 75005. Salle de réunion, RDC


Abstract :

The current philosophical debate on memory is dominated by two camps. On one side, we face modified versions of the Causal Theory that hold on to the idea that remembering requires a memory trace that causally links the event of remembering to the event of perception and carries over representational content from the content of perception to the content of remembering. On the other side, a new camp of Simulationists is currently forming up. Motivated by empirical and conceptual deficits of the Causal Theory and its modifications, they reject, both, the necessity of preserving representational content and the necessity of a causal link between perception and memory. They argue that remembering is nothing, but a specific form of imagination, and differs from other forms only in that it has been reliably produced and is directed towards an episode of one’s personal past. Sharing the criticism of the Causal Theory and its demand for an intermediary carrier of representational content, I will argue that a causal connection to experience is, still, necessary to fulfill even
the minimal requirements of past-directedness and reliability, accepted even by Simulationists. I will develop an account of minimal traces devoid of representational content and exploit an analogy to the predictive processing framework of perception. As perception can be regarded a prediction of the present on the basis of sparse sensory inputs without any representational content, episodic memory can be conceived of as a “prediction of the past” on the basis of a merely causal link to a previous experience. The resulting notion of episodic memory will be validated as a natural kind distinct from mere imaginary processes.

 

 


 

 


EHESSCNRSENS