@JFRamadier Profile picture

JF Ramadier

@JFRamadier

Entrepeneur #digitalhealth @ClinePlatform #softwareengineer #cto #coding #machinelearning

Similar User
LauMa communication photo

@laumacom

Catherine de Rohan photo

@cderohan

ETIENNE Loic photo

@drLoicETIENNE

Pierre-Yves Arnoux photo

@pierreyves

Julien Livis photo

@JulienLivis

Basil Strategies photo

@Basilstrategies

Delphine Eriau photo

@DelScope

Lucas Freton photo

@LFreton

Interesting!

K-nearest neighbor can provably learn any function given enough data and compute. So if scaling is all you need, AI was solved in 1951.



JF Ramadier Reposted

A data scientist isn't a scientist, a prompt engineer isn't an engineer, machine learning doesn't learn, artificial intelligence isn't intelligence, and agents don't have agency. This industry is so full of shit.


💯

A lot of automation technology doesn't actually save any labor once you factor in all of the interactions you have with it. A drip coffee maker would be a canonical example.



😂👍

your startup is just an openai wrapper openai is just an nvidia wrapper nvidia is just a tsmc wrapper tsmc is just an asml wrapper asml is just a sand wrapper



JF Ramadier Reposted

Relatable. I debug all my Python code with tactical print() drops

The secret behind my productivity is that I'm almost exclusively a print-statement debugger



JF Ramadier Reposted

I’m not interested in studying LLMs. I’m interested in making them obsolete.


💯 d’accord! 👍

Oui 100% d’accord c’est bien pour les fonction de base et les libe et encore il faut savoir les inclure dans son code



💯👍

Software engineering is only 20% coding. The remaining 80% is problem solving, resourcefulness and creativity LLMs can do a reasonable job of coding! It’s the 80% that they are behind on….



JF Ramadier Reposted

20% des actes médicaux ne seraient pas pertinents. Une économie potentielle de 250mds... Si l'on n'avait pas une culture aussi conservatrice à l'égard des données de santé, et totalement décorrelé risque et opportunité, des outils de Big data /ml permettraient facilement…


JF Ramadier Reposted

Physicists think AI is physics. Statisticians think AI is statistics. Mathematicians think AI is mathematics. Psychologists think AI is psychology. Neuroscientists think AI is neuroscience. And they’re all right.


JF Ramadier Reposted

LLMs have 1-2 years of relevance left.


JF Ramadier Reposted

1 logistic regression in production is worth more than 2,332 neural networks in development


JF Ramadier Reposted

l'Union Européenne, dont le retard est déjà significatif, pourrait rater la révolution IA, si elle continue de réguler celle ci et poser des contraintes trop fortes sur les données d'apprentissage. ericsson.com/en/news/2024/9…


JF Ramadier Reposted

Because it's a language model, 10000 and 100000 are much closer in the latent space than they should be, while 10000 and 9999 are much farther apart than should be.


JF Ramadier Reposted

The AI era is dawning, but the sun hasn’t risen yet.


JF Ramadier Reposted

Every step forward counts.

Tweet Image 1

👍 Programming is 1% writing code and 99% design thinking! Are you agree @svpino?

Programming is 1% writing code and 99% understanding why the code you wrote is not working.



JF Ramadier Reposted

La complexité de la pensée n’est pas atteignable par les IA sans corps et sans appétit de vie ou conatus (spinoza). C’est de l’imitation de surface, des productions grâce à des mécanismes statistiques sans intention

Si la question de la conscience n'a pas de sens pour les actuelles IA, il faudra sans doute réfléchir bientôt à la définition de l'intelligence. Rien ne nous dit que nos propres phrases/pensées ne sont pas -comme pour l'IA- probabilistes afin de répondre au mieux à un objectif



Ca c’est clair ! Y en a-t-il qui pense le contraire @lau_devil !?

L’IA generative modélise le langage à partir des régularités des données d’apprentissage sans modéliser le processus de la pensée ! Ce n’est pas avec plus de données que cela changera la capacité d’apprentissage du processus de pensée ! Il faut d’autres algos d’IA !



JF Ramadier Reposted

Neural networks are ten times simpler than ordinary software, but it's ten times harder to make them work.


Loading...

Something went wrong.


Something went wrong.