There is a big flaw in asking an AI of 2025 to say what are the chances of AI to "this or that" in the future, as if AI was thinking to it in some way. It isn't. AI of 2025 is still trained from scratch on human made scientific papers and human made ideas, that the AI of 2025 will spit out in a "I've seen this somewhere" kind of way. So, if you're scared of something and ask an AI (which is an echo of you) about its opinion on the matter, you may end up in a Larsen kind of situation where you exacerbate the fear that you already have by receiving your fear but in a different form that you're used to. And I insisted a lot on "AI of 2025" because we don't know yet if in the near future or in future, there will be a different kind of AI that will think for itself, meaning that it would NOT be trained on human knowledge and information. That would be a shift like the one from Alpha Go to Alpha Go Zero (from Google). It's theoretically possible, but hasn't been done. "Experts AI disaster scenario just came true" with Stephen Fry - Pindex

"AGI is humanity's next big moment and we have companies like OpenAI, Anthropic, xAI, DeepSeek, MoonShot, Alibaba, AntGroup and more, all trying to achieve AGI to some extent. But how have we been approaching this pursuit and what is humanity's cost to achieve AGI?"

Why most AI startups are bleeding al lot of money just to exist and to run, and are not even close to break even by TechButMakeItReal

"How Much Has The World Spent on AI?... So Far" by How Money Works

After Skool - Is AI Apocalypse Inevitable? by Tristan Harris

a very interesting conference about testing general intelligence of models

When it comes to AI, the time is nigh 🤺 "Why AI experts say humans have two years left. Stephen Fry"

Google Gemini 2.5 Pro is becoming crazy good and impressive, especially in coding:

An AI to change the world?

IBM utilise déjà des intelligences artificielles pour déterminer qui de ses employés va vouloir démissionner bientôt, afin de l'anticiper et l'éviter. C'est brillant !

"Les 6 dangers de l'IA selon le MIT : Les voitures autonomesLa manipulation politiqueLes algorithmes tueursLa reconnaissance facialeL'ère du deep fakeLa discrimination algorithmique"

L'intelligence artificielle n'est pas du tout ce qu'on croit. Pour la plupart d'entre elles, elle ne fonctionne que comme un miroir, un écho de la réalité. Et si elle s'occupe de générer un écho d'intelligence humaine, alors les gens vont immédiatement penser que c'est elle qui est la source de l'intelligence. Or rien n'est moins vrai. Une intelligence artificielle n'est pas plus intelligente qu'une chaîne HiFi n'est capable de parler ou de jouer de la musique, ou encore qu'un appareil photo n'est capable de dessiner un paysage. L'intelligence artificielle est pratique, certes, parce qu'elle automatise grandement des processus qui peuvent nous êtres utiles. Mais une intelligence humaine sera toujours nécessaire pour déterminer ce qui est utile.