Artificial Intelligence - Jacques Pitrat French pioneer

Pioneer in Strong Artificial Intelligence

Jacques Pitrat passed away on 14 October 2019. Graduate of the X, Engineer of the Armament, Director of Research at the CNRS, Jacques Pitrat devoted his life to promoting and encouraging the development of Artificial Intelligence in France.

Pioneer of Artificial Intelligence and author of numerous publications, father of Artificial Intelligence in France and trainer of many colleagues in the field of AI, active member of the French Association for Artificial Intelligence, it is to this generous man, admirable and respected by all, that the national AI community wishes to pay tribute to him through numerous testimonies.

Find all the interventions of the participants in the form of videos: Tribute to Jacques Pitrat in Videos!

"Jacques Pitrat: Director of Research, Director of Artificial Intelligence Teams

by Monique Grandbastien (Lorraine University) & Monique Baron (Sorbonne University)

Four periods of Artificial Intelligence at Jussieu with Jacques Pitrat

1. Start-up, the time of the pioneers (1967-1977) - Institute of Prog., GR22
2. Pitrat-Laurière" period (1977-1989) - GR22-CF Picard, then LAFORIA
3. Equipe Métaconnaissances - Jacques Pitrat continued (1989-1997) - LAFORIA
4. LIP6 period (from 1997)

Contents :

• Artificial Intelligence training in postgraduate studies, DEA & thesis supervision;
• Team life ;
• Development/evolution of research topics;
• Contributions to the development and influence of Artificial Intelligence in France and beyond.

Jacques Pitrat has had a busy career:

• Over 35 years of training and supervision of theses in Artificial Intelligence
• More than 60 years of research in Artificial Intelligence.

From 1967 to the first decade of the new millennium, Jacques Pitrat has trained in Artificial Intelligence and accompanied in research several generations of students and researchers (70 theses), with very great scientific and human qualities unanimously recognised and appreciated.

Some of his "disciples" have continued in this way, including:

• C Lemaître in Mexico, MF Verdejo and F Garijoen Spain .../...
• Anne Adam (Caen), JP Laurent (Chambéry), M Vivet(Le Mans), D Pastre(U Paris-Descartes), MO Cordier (Orsay, Rennes), JM Fouet and N Guin(Lyon), M Grandbastien(Nancy), B Bouzy and Y Parchemal(U Paris-Descartes), S Pinson and TCazenave(U Paris-Dauphine), J Courtois (EPITA) .../... [and the following generations]

"Knowledge, Chocolate Béchamel"

by Marie-Odile Cordier (Rennes 1 University)

Controlling a robot in natural language in a domain requiring pragmatic knowledge: cooking recipes.

Natural language assumes the existence of a context. It is the set of sentence and context that contains the information. The sentence alone is a priori ambiguous. It is the user's knowledge that removes ambiguities. The objective of conciseness is to minimise the sentence while remaining understandable.

It is therefore necessary to transmit to the program the knowledge assumed by the speaker about the context.

The languages used to describe this knowledge must be easy to use to ensure adaptation to other contexts. Here procedural languages for DICO1 and DICO2...

Robot control type language.

"Automatic demonstration, metatheorems, declarative knowledge"

by Dominique Pastre (Paris Descartes University Honorary)

In 1966, Jacques Pitrat published a first thesis on : "Realization of theorem proving programs using heuristic methods".

The aim of the work was to write a general programmae that discovers and/or proves theorems in various axiomatics of propositional logic.

"Jacques Pitrat - Artificial Intelligence and games"

by Tristan Cazenave (Paris-Dauphine University)

Feedback on the thesis with Jacues Pitrat on the theme: "Self-observing learning system - Application to the game of Go".

Jacques Pitrat's original ideas on Artificial Intelligence and games, Bootstrap and General Game Playing, are at the heart of today's game systems that perform much better than humans. The applications go far beyond games.

"Taking into account symmetries in CAIA"

by Jean-Yves Lucas (EDF Saclay)

Discovering symmetries is essential to reduce combinatorics, or to prove properties of the problem to be solved (e.g. no unique solution...).

Symmetries are extremely useful to solve combinatorial problems intelligently.

"Jacques Pitrat, meta-knowledge and the AI Bootstrap"

by Marc Porcheron (EDF Saclay)

"Systems with superior intelligence must be able to know the domain in which they work, but also to examine what they should do, what they can do, what they know, what they are doing.

They operate at a 'meta' level, above the level of the problem at hand; knowledge becomes for them an object of study.

Metaknowledge is precisely the tool that allows us to work on knowledge and to create such highly intelligent systems."

Jacques Pitrat showed that these principles could be applied to the bootstrapping of metaknowledge, using a "start-up engine".

Marc Porcheron, a senior engineer at EDF R&D Saclay, looks back at his thesis directed by Jacques Pitrat in 1990, dealing with a meta-expertise for compiling languages based on production rules.

"Consciousness and Artificial Intelligence seen by Jacques Pitrat"

by Gérard Sabah (CNRS, honorary research director)

Why are we interested in consciousness?

Edelman: "The functionalities necessary for true intelligence are those which, based on the unconscious, allow the emergence of consciousness in humans.

Why not in machines?

Gérard Sabah (CNRS, honorary director of research) returned to some basic notions concerning Artificial Intelligence put forward by Jacques Pitrat:

• Artificial Intelligence is the most difficult problem that man has tackled.
• Man may not be intelligent enough to solve it.
• Artificial Intelligence systems themselves need to be helped with priming.
• META: After solving the problem, generalise and learn.
• One does not copy human intelligence but it is often good to learn from it.
• Artificial cognition ≠ human cognition ;Some artificial cognitive abilities are inaccessible to humans (and vice versa
• Let's not forget strong Artificial Intelligence!

« From CAIA... To refpersys reflexive, introspective, meta-basedAI systems »

par Basile Starynkevitch

• What is AI ?
• Why AI systems are needed? hard problems.
• Why AI systems need to be free software?
• Pitrat’s thesis about intelligence and consequences.
• Engineering aspects of AI software systems
• What should really matter in an AI system?
• AI software and computers “are” semantics networks
• The staircase development model of bootstrapped AI.
• CAIA in practice.
• RefPerSys - a future successor to CAIA.

"The scientific work of Jacques Pitrat (1934-2019), a historical perspective"

by Jean-Paul Haton (Lorraine University) & Henri Prade (CNRS Toulouse)

• Jacques Pitrat a pioneer of Artificial Intelligence ... symbolic in France.
• He revealed a great continuity and coherence in his research.
• His research was driven by a very ambitious vision of the future of Artificial Intelligence.
• Jacques Pitrat was a supporter of strong Artificial Intelligence (cf. his presentation a year ago hnps://www.lip6.fr/colloquium/?guest=Pitrat).
• Jacques Pitrat was a great trainer in his discipline with the supervision of 70 theses (all in Artificial Intelligence).
• He was the first Frenchman to be a member of the Editorial Committee of "Artificial Intelligence".
• He was an EuRAI, an AAAI fellow.
• Defender of French Artificial Intelligence.
• Programmer and experimenter to the end!
• A simple, charming and passionate man.

"Jacques Pitrat, the man who could sew on a button"

by Michel Buthion (Retired from Transports Publics Lausannois)

Artificial Intelligence

Why

A chess game started at the beginning of mankind between two computers that would be ten times more powerful than the most powerful that exists today and would develop the complete tree of possible moves would today be on the 2nd move only.

Limit

If I tell a computer "I'm flying tomorrow", it will never say "so get the sewing machine out". But a human being can do it: a button on my coat has been undone for several days and if I go on a trip, I can't wait to sew it up.

Michel Buthion recalls his doctoral thesis, directed by Jacques Pitrat and defended in 1975, on the subject: A program that formally solves geometric construction problems.

"Jacques Pitrat: the man as I met him"

by Joël Courtois (Epita)

Joël Courtois presents SIAM or Système Intelligent pour l'Assistance aux Manipulations, a diagnostic system that can be easily adapted to new fields and that teaches his method (1990).

For Joël Courtois, Jacques Pitrat was an outstanding Research Director and reiterates his amazement at his memory and power of analysis.

Jacques Pitrat was a visionary at the service of science and humanity, with an astonishing kindness and patience.

"Was it a good idea to fund the Pitrat team?"

by Jean-Luc Dormoy (@jldormoy)

The bootstrap

Humans are too limited to build an AGI alone.

Programming is too hard

Machine Learning (e.g. Go):

• The limitations of humans in problem solving are visible.
• The limitations of humans in programming are even more visible

General systems design :

• Organ and function: apply one to the other and the other to the one

Consciousness

You are the object of my thoughts

Checking is too hard

Machine Learning (e.g. Autonomous Car):

• Human limitations in driving safe systems are visible.
• Human limitations in designing safe systems even more visible.

General system design :

• Leave verification to operation.
• (Learning verification).

Next steps in AI

What needs to be funded?

"1970-1972: bootstrapping AI with Jacques Pitrat"

by Claude Roche (Consultant)

Closing of the day by Yves Demazeau (President of AfIA - CNRS)