🚀Best evidence yet of alien life, prehistoric wolf resurrected, the 4 ways AGI may destroy us, talking to dolphins & more
"The only scenario that currently explains all the data is an ocean world teeming with life"
Bonjour,
Vous recevez la version gratuite de la newsletter Parlons Futur : une fois par semaine au plus, une sélection de news, mêlant sources anglophones et francophones, résumées en bullet points sur des sujets tech 🤖, science 🔬, éco 💰, géopolitique 🌏 et défense ⚔️ pour mieux appréhender le futur 🔮.
Je m'appelle Thomas, co-fondateur de YeldaAI, qui développe des IA pour répondre au téléphone en langage humain pour les administrations et les entreprises. Plus d'infos sur moi en cliquant ici.
Mes derniers podcasts
Entretien avec un expert défense : fin de la guerre en Ukraine ; Taiwan, cause perdue ? ; guerre et IA, la révolution (Youtube, Spotify, Apple)
Entretien exclusif avec Philippe Bihouix, co-auteur de la BD à succès "Ressources : Un défi pour l'humanité" (Youtube, Spotify, Apple)
Vous l'avez peut-être déjà entendu intervenir aux côté de Jean-Marc Jancovici sur Thinkerview ou ailleurs, Philippe Bihouix est passionnant !
100 ans de progrès compressés en 10 ans ? Le débat qui agite la haute sphère de l'IA (Youtube, Spotify, Apple)
Et toujours : entretien avec Jacques Attali (YT, Apple, Spotify), avec Pierre Bellanger, fondateur et CEO du groupe Skyrock sur Apple Podcast et Spotify).
Et voici donc ma toute dernière sélection !
Wow: Scientists find strongest evidence yet of life on an alien planet (NYT)
A team of researchers is offering what it contends is the strongest indication yet of extraterrestrial life, not in our solar system but on a massive planet, known as K2-18b, that orbits a star 120 light-years from Earth. A repeated analysis of the exoplanet’s atmosphere suggests an abundance of a molecule (dimethyl sulfide or DMS) that on Earth has only one known source: living organisms such as marine algae.
"The only scenario that currently explains all the data obtained so far from JWST (James Webb Space Telescope), including the past and present observations, is one where K2-18 b is an ocean world teeming with life"
“We spent an enormous amount of time just trying to get rid of the signal.”
No matter how the scientists revisited their readings, the signal stayed strong.
"Not yet a biosignature, but a timely prompt to sharpen our methods and expectations," concluded astrophysicist Sara Seager, who was also not involved, told the Washington Post.
Have you heard of the Enhanced Games? In which athletes can earn $1m for breaking world records using the sorts of doping that would normally disqualify them.
Meet the new Kawasaki CORLEO: a four-legged robot that humans can ride (vidéo de 2 minutes)
Prototype décoiffant…en images de synthèse pour l’instant…
Dans le même genre, images réelles, mais prototype bien lent: This Rideable Rhinoceros Robot Looks Absolutely Amazing
Startup Earth AI’s algorithms are discovering overlooked deposits of copper, cobalt, and gold in Australia.
With AI, a woman with complete paralysis and zero vocal function just regained the ability to speak at ~conversational speed. (Nature, X thread)
a brain implant with only a couple hundred sensors learned to read her intention to speak and synthesized her words in her own voice, in real time, not typing or text, actual speech.
before) this, the best systems let patients “type” at 8–14 words per minute (WPM).
this one does 90+ WPM. in audio, with prosody. from brain activity alone.
and it uses no audible training data. she doesn’t even need to try making sounds.this will enable silent communication and shows that language can be purely neural
In a striking recent example, Chinese electric-vehicle maker BYD, which just surpassed $100 billion in revenues, is reportedly delaying its planned $1 billion Mexican factory over concerns that its advanced technology could leak to US rivals! (Financial Times)
Chinese regulators grant the country's first-ever approvals for two companies to operate autonomous passenger drones (flying taxis). (source)
China ended up being the first to invent the flying car, because they mastered battery tech and electric motors before we did.
One shipyard in China made more commercial ships last year than the total number the U.S. has produced since World War Two. (source)
NYT's journalist Tom Friedman reporting on his trip to China:
I’d never seen anything like this Huawei campus. Built in just over three years, it consists of 104 individually designed buildings, connected by a Disney-like monorail, housing labs for up to 35,000 scientists, engineers and other workers, offering 100 cafes (…)
Huawei also went into the business of creating the A.I. technology for everything from electric vehicles, self-driving cars and even autonomous mining equipment that can replace human miners. Huawei officials said in 2024 alone it installed 100,000 fast chargers across China for its electric vehicles; by contrast, in 2021 the U.S. Congress allocated $7.5 billion toward a network of charging stations, but as of November this network had only 214 operational chargers across 12 states.
What makes China’s manufacturing juggernaut so powerful today is not that it just makes things cheaper; it makes them cheaper, faster, better, smarter and increasingly infused with A.I.
Security researchers have uncovered a pre-installed, undocumented remote access tunnel in ChineseUnitree Go1 robot dogs. (source)
This raises serious concerns about supply chain trust, especially as these robots are widely used in academic, corporate, and even defense-related environments.
Palmer Luckey, founder & CEO of Anduril, an American defense technology company that specializes in autonomous systems, earned a standing ovation at TED last week (source):
“Your Tesla has better AI than any military aircraft.”
“Countries only go to war when they disagree who the victor will be.”
“We have to prove we have the capacity to win.”
“Manufacturing capacity is our greatest vulnerability. We need more manufacturing and smarter weapons. AI is the only possible way we can keep up with China’s numerical advantages.”
AI making the kill decision: “I love killer robots.” (Prelude to his more serious answer that we have been doing this already for many years, and it’s the inevitable future).
Ma tribune sur le sujet dans le JDN: Débat sur les armes autonomes : la liste des arguments pour et contre
Drones are now more lethal in Ukraine (70% of deaths) than artillery was in the WW2 (65% of deaths), the most artillery heavy war in human history to date.
GPT-4.5 et LLaMa-3.1-405B ont réussi le test de Turing (scientific paper, trouvé par le JDN)
75 years ago, Alan Turing (1950) proposed the imitation game as a method of determining whether machines could be said to be intelligent. In the game—now widely known as the Turing test—a human interrogator speaks simultaneously to two witnesses (one human and one machine) via a text-only interface. Both witnesses attempt to persuade the interrogator that they are the real human. If the interrogator cannot reliably identify the human, the machine is said to have passed: an indication of its ability to imitate humanlike intelligence
"People were no better than chance at distinguishing humans from GPT-4.5 and LLaMa (with the persona prompt)," wrote lead author
GPT-4.5 a été jugé humain par les interrogateurs dans 73% des cas. Un taux significativement supérieur à celui des véritables participants humains.
...but the AI was specifically told to put on a specific persona, like a young person who's knowledgeable about the internet and culture.
Without persona prompting, GPT-4.5 achieved an overall win rate of merely 36%
LLMs seem to have a universal "language of thought" shared across languages, it doesn't just memorize translations but processes concepts in a language-agnostic space. (The Economist)
When a chatbot is multilingual, is there in effect an entire second copy of everything it knows, or does it have some awareness of concepts that transcend language?
Ask it in English for the opposite to “big”, in French for the opposite to “grand” or in Chinese for the opposite to the Hanzi character for the same concept, and the same feature lights up in every case, before more language-specific circuits kick in to “translate” the concept of smallness into a particular word.
Shopify CEO says no new hires without proof AI can’t do the job (The Verge)
Scary: ew research shows Reasoning models don't accurately verbalize their reasoning (Anthropic AI)
So monitoring their verbalized steps (their Chain of Thoughts) is unlikely to reliably catch rare, catastrophic behaviors
DeepMind has detailed all the ways AGI could wreck the world (ArsTechnica)
Misuse: The user instructs the AI system to cause harm
Misalignment: The AI system takes actions that it knows the developer didn’t intend
Mistakes: The AI system causes harm without realizing it
Structural risks: Harms from multi-agent dynamics, where no single agent is at fault
In a study with 469 teachers in Sierra Leone, we find AI works better than the web--and is 87% cheaper. (source)
because some AI chatbots can be hosted on a device and be used without internet
Only 37% of sub-Saharan Africans use the internet. Cost is the #1 constraint.
Compilation of 16 interesting types of videos and images people did with GPT 4o native image generation can do (aside from Ghibli images) (source)
including: Full Lord of the Rings video trailer in Ghibli, Videos in different animation styles, Video ads, converting famous images into Voxel Art, interior decor, Stylized infographic maps!, comic strips, Create 3D models from images with depth maps! and more
5 short video examples of what Runway's latest state-of-the-art AI model can do for media generation and world consistency: watch the videos
ChatGPT’s new image generator is really good at faking receipts (TechCrunch)
Lol: A man who used an AI avatar in court because he thought it would present an argument well says he got chewed out by a panel of judges (Fortune)
It took only seconds for the judges on a New York appeals court to realize that the man addressing them from a video screen — a person about to present an argument in a lawsuit — not only had no law degree, but didn’t exist at all.
"May it please the court,” the man began. “I come here today a humble pro se before a panel of five distinguished justices.”
“Ok, hold on,” Manzanet-Daniels said. “Is that counsel for the case?”
“I generated that. That’s not a real person,” Dewald answered.
It was, in fact, an avatar generated by artificial intelligence. The judge was not pleased.
Bonobos’ Calls May Be the Closest Thing to Animal Language We’ve Seen (ArsTechnica)
"A team of Swiss scientists led by Melissa Berthet, an evolutionary anthropologist at the University of Zurich, discovered bonobos can combine [vocal calls including peeps, hoots, yelps, grunts, and whistles] into larger semantic structures. In these communications, meaning is something more than just a sum of individual calls—a trait known as non-trivial compositionality, which we once thought was uniquely human."
For instance, the “I would like to” peep followed by “let’s stay together” whistle was used for initiating more romantically inclined interactions bonobos are famous for indulging in.
DolphinGemma: Google AI is helping decode dolphin communication (source)
It led, among other things, to an underwater computer designed not to directly decipher the dolphins' complex natural language, but to establish a simpler, shared vocabulary. Watch the 1-min video
Partagez cette newsletter par Whatsapp en cliquant ici ❤️
Si vous appréciez cette première moitité de l’apéro gratuite, n’hésitez pas à prendre 3 secondes svp pour transférer cet email ou l’envoyer à ne serait-ce qu’un contact par Whatsapp en cliquant ici 🙂
Et si on vous a partagé cet email, vous pouvez cliquer ici pour vous inscrire et ne pas manquer les prochains
Meet Romulus and Remus—the world’s first de-extinct animals, born on October 1, 2024. (CNN)
Dire wolf : un loup préhistorique à l'énorme mâchoire
The dire wolf has been extinct for over 10,000 years. These two wolves were brought back from extinction using the CRISPR technique to make genetic edits derived from a complete dire wolf genome, meticulously reconstructed by Colossal from ancient DNA found in fossils dating back 11,500 and 72,000 years.
“Our team took DNA from a 13,000 year old tooth and a 72,000 year old skull and made healthy dire wolf puppies.”
You can hear the first howl of a dire wolf in over 10,000 years here
Colossal scientists have created three dire wolf pups by using ancient DNA, cloning and gene-editing technology to alter the genes of a gray wolf, the prehistoric dire wolf’s closest living relative, the company announced Monday. The result is essentially a hybrid species similar in appearance to its extinct forerunner.
Bon : There’s no secret that across the genome, this is actually 99.9% gray wolf (le loup gris bien viant aujourd'hui). There is going to be an argument in the scientific community regarding how many genes need to be changed to make a dire wolf, but this is really a philosophical question,” Dalén said.
Dire wolves and gray wolves shared 99.5% of their DNA anyway, according to the company’s news release.
“We aren’t trying to bring something back that’s 100% genetically identical to another species. Our goal with de-extinction is always create functional copies of these extinct species.
To achieve its goal, the company essentially created a hybrid genome using CRISPR technology to cut away certain gray wolf gene variants and replace them with traits associated with dire wolves
“It carries dire wolf genes, and these genes make it look more like a dire wolf than anything we’ve seen in the last 13,000 years. And that is very cool.”
Are phones are making us dumb? (Financial Times)
Recent data show a decline in cognitive skills in rich countries that coincides uncomfortably with the rise of the smartphone. Here’s a chart that shows a decline in the U.S. that starts around 2012
Another chart suggests the decline is a global phenomenon
The mechanism here isn’t hard to deduce: A social-media-enabled smartphone is a distraction machine that barrages our mind with constant notifications and tiny bits of unrelated information, destroying our ability to focus and think carefully about anything.
A third chart lends weight to this hypothesis, showing a steep rise in the % of people struggling to concentrate or learn new things.
Modern China is certainly a very innovative country. Chinese scientists now publish the majority of high-impact papers in fields like chemistry, physics, computer science, materials science, and engineering (The Economist):
The country’s true dominance is probably less than depicted in this chart, due to “home bias” in the citations used to measure papers’ impact. But even correcting for that bias, China is undeniably a scientific superpower.
3 cool and tangible examples of how AI models are helping dirty industries go green (The Economist)
Mining is another dirty business where ai is making inroads. Fortescue, an Australian giant, is applying ai in designing current systems and redesigning future mining and energy operations with an eye to eliminating fossil fuels. Its algorithms automate tasks such as calculating how energy is used and the routes that autonomous heavy vehicles take. If the weather forecast is for rain, meaning solar output will fall, the company brings forward energy-intensive tasks while it can still use clean solar power. The software enabling this sort of load flexibility, the firm reckons, has allowed it to cut the required capacity of the power system it built by 9%, saving nearly $500m.
Shipping and logistics companies have taken to applying ai with gusto. ups, a package-delivery giant, recalculates delivery routes throughout the day as orders, pickups and traffic conditions fluctuate. It estimates that its smart software has improved on-time delivery while cutting 16-22 kilometres from drivers’ daily trips, saving hundreds of millions in fuel costs.
Denmark’s Maersk, one of the world’s largest container-shipping lines, uses ai to analyse variables from engine performance to ocean currents and weather to avoid rough seas and waiting time. Making even its older ships smarter has reduced fuel consumption by over 5% across its fleet, saving $250m and reducing CO₂ emissions by perhaps 1.5m tonnes.
👋👀🦉💡Je vous invite maintenant à vous abonner à la version payante (l’équivalent du prix d’un café par semaine, le minium imposé par Substack) pour recevoir la semaine prochaine les résumés de ces longs articles issus des plus grands journaux (et accéder aux futures éditions payantes et archives) en cliquant sur le bouton “Upgrade to paid/Subscribe now” 👇
🔒Pig hearts and kidneys have been transplanted into humans with mixed results, now find out what organ Chinese scientists managed to transplant
🔒This is the first fully AI-generated paper to pass the same peer-review process as human scientists
🔒Nouveau concept : "vibecoding", Not a Coder? With A.I., Just Having an Idea Can Be Enough.
🔒Important: In case of war with China, the US will need a strong manufacturing sector, see how badly it’s losing
Les dernières newsletters :
Merci, et à bientôt !
Thomas