👽 Fermi paradox update; conquering Taiwan in 3 hours?; see who won that robot-vs-human manipulation contest

See that huge driverless Chinese mining truck & more

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 🌏, défense ⚔️ et espace 🚀 pour mieux appréhender le futur 🔮.

Je m’appelle Thomas, plus d’infos sur moi en cliquant ici.

🎙️ Mes derniers podcasts

• Poutine et la guerre à l’Europe : le scénario catastrophe du front atlantique

  • Entretien exceptionnel (Spotify, Apple, Youtube) avec Stéphane Audrand, consultant indépendant en risques internationaux, historien et officier de réserve dans la Marine. Il est l’auteur remarqué de l’article “Poutine et la guerre à l’Europe : le scénario du front atlantique” publié dans le Grand Continent.

• Vivons-nous dans une simulation ? (Youtube)

• 🌍Entretien exclusif avec Philippe Bihouix, co-auteur de la BD à succès “Ressources : Un défi pour l’humanité” (Youtube, Spotify, Apple)

Et voici donc ma toute dernière sélection !

• Passionnant : Arthur Mensch, cofondateur de Mistral AI, auditionné à l’Assemblée nationale (audition de 1h20 sur Youtube)

  • 3 grands impacts:

  • Impact sur l’emploi : L’IA devrait remplacer 10% des emplois assez vite, d’après lui !

  • Une course de vitesse face aux Américains : l’IA cannibalise l’offre énergétique, et les premiers arrivés seront les premiers servis, il faut donc que les entreprises européennes construisent infra énergétiques et centres de données au plus vite pour que les autres ne le fassent pas avant, et qu’on ne dépende pas davantage d’eux

  • Enjeu de souveraineté : 10% de la masse salariale d’aujourd’hui serait donc bientôt dépensée en IA très bientôt, soit des centaines de milliards d’euros à payer aux Etats-Unis si on ne peut pas se fournir localement

    • il faut compter 1kW de puissance requise pour générer l’IA permettant de “remplacer” un salarié (dont le travail peut l’être...), nécessitant la puissance de calcul d’une demi puce GPU, coût estimé à 10k€ par an

    • 400 millions de personnes * 10% dont le travail devrait être fait par l’IA à horizon 5-10 ans* 10,000€/an = autour de 400 milliards €

    • ça en moins dont les profits auraient pu être investis localement en R&D

• Was the China shock on employment exaggerated? (The Economist)

  • Perhaps 2m Americans lost their jobs between 1999 and 2011 owing to China’s entry into the global trading system.

  • That is no worse than a typical month’s lay-offs in America’s churning labour market

• China has tested a directed-energy beam that recharges drones mid-flight. (source)

  • A step towards 24/7 flying drones

• Russians Fear Ukraine ‘Slaughterbot’ Drones Are Head-Hunting Them (Forbes)

  • Russian military bloggers are warning of a new type of Ukrainian FPV drone with thermal imaging and AI which detects the target’s face and fires a high-velocity projectile at them. A video shows a Russian soldier cut down with what look like a precise head shot from one such drone. A further video confirms the kill.

  • Drones with EFP warheads are an obvious countermeasure to netting, wire cages and other forms of protection including “turtle tanks” as the drone can attack from some distance and the slug will go right through such protection. However, this type of warhead needs to be aimed and detonated at the right moment which requires a higher level of operator skill.

• Putin’s failure (Ian Bremmer)

  • The country that was supposed to fall in three days, then three weeks, has become the most capable drone-warfare force on Earth and is fast approaching full military self-sufficiency – with US weapons now accounting for only about 20% of Ukraine’s arsenal and declining.

  • Putin set out to erase Ukraine from the map and weaken Europe. He may have forged the world’s most formidable new military-technological power instead – and in doing so, handed Europe its most important long-term strategic asset: a battle-hardened, drone-manufacturing, AI-forward Ukraine that is increasingly capable of defending itself and the continent.

  • That’s a big win for Ukraine. It’s a big win for Europe. And it’s a big win for democracy and rule of law, despite all of Ukraine’s very real corruption and institutional inadequacy – none of which comes close to what we see in Russia.

• Clash : between a German tank maker and a Ukraine drone (source)

  • The CEO of Rheinmetall, recently sort of ridiculed [the] Ukrainian drone industry, saying that…there is nothing interesting there, no real innovation…

  • Alexey Babenko, head of and founder of VIARI Drone, a Ukrainian company which is one of the largest manufacturers of FPV drones said that :

    • the drones we manufacture in one day will be more than enough to destroy all the tanks Rheinmetall manufactures in a year…

  • Cost-wise, of course, a drone is like, $500 and a Rheinmetall tank is what, probably 5 million-ish or maybe more…

• Will China conquer Taiwan in a few hours with its drones? (source)

  • Yaroslav Azhnyuk, founder and CEO of The Fourth Law, one of Ukraine’s most important drone startups said :

    • Last year, Ukraine produced 4 million FPV drones. Ukraine is not the most industrious nation in the world.

    • China can produce 4 billion of these FPV drones…China can [also] make…fixed-wing drones, which go not forty kilometers far, but maybe two to three hundred kilometers inland…

    • They can also make them all fully autonomous. They have DJI, the world’s most advanced drone company. They can make them fully autonomous without GPS, without anything.

    • Then they can put those drones on maybe tens of thousands of fully autonomous underwater submarines, or maybe not even that just on shipping containers and barges that ship goods or freight ships. And then they show up with millions of drones packed onto those sea vessels.

    • They show up to any coastline in the world, be it Taiwan or be it California, and they have millions of long-range impactors targeted at a piece of land.

• One Polymarket trader made $21,000 by tampering with the one weather sensor at Charles de Gaulle Airport used to settle contracts on Paris temperature. (source)

• Uber COO not sure its AI costs are worth it (source)

  • “I think maybe implicitly there is more that is getting shipped, but it’s very hard to draw a line between one of those stats and, ‘Okay, now we’re actually producing 25 percent more useful consumer features.’”

  • “If you’re not actually able to draw a direct line to how much useful features and functionality you’re shipping to your users that trade becomes harder to justify because it’s not free,”

• Crazy but true apparently: More Europeans die of summer heat than Americans die of guns (Hannah Ritchie)

  • true in absolute numbers but not per capital

• SpaceX’s target valuation of around $1.75trn is more than 90 times the $18.7bn in revenue the firm brought in last year. (The Economist)

  • Tesla, Mr Musk’s electric-vehicle company, trades at a mere 16 times revenue.

• Morgan Stanley, an investment bank, estimates that agentic systems require one CPU for every GPU, compared with a ratio of one to 12 for chatbot-style systems.

  • A CPU is designed to handle complex, step-by-step tasks sequentially. Originally the brain of smartphones or computers.

  • A GPU is designed to handle thousands of simple tasks at once (parallel processing), making it perfect for graphics, gaming, video editing, and AI.

  • While the first era of AI was all about GPUs (making Nvdia fortunes), now with agentic AI (AI that does stuff, not just answer queries), we’re rebalancing to CPUs

• OpenAI co-founder Andrej Karpathy joins Anthropic’s pre-training team (Techcrunch)

• First AI Model From Zuckerberg’s Wildly Expensive Superintelligence Lab Flops Compared to Virtually All Rivals

  • An executive told Bloomberg that the new model won’t be able to keep up with OpenAI’s ChatGPT, Anthropic’s Claude or Google’s Gemini.

  • In a blog post announcing the new model, the company admitted Muse Spark “is an early data point on our trajectory, and we have larger models in development.

• With the help of Claude Mythos Preview, the Firefox team fixed more security bugs in April than in the past 15 months combined. (source)

• “C’est le calme avant la tempête” : Google alerte sur l’industrialisation des cyberattaques par IA (JDN)

  • Les cas documentés par Google Threat Intelligence sont assez sidérants. L’exemple le plus marquant ? Un malware déployé par le GRU, le renseignement militaire russe. Particularité, il n’embarque pas ses commandes en dur. Au moment de s’exécuter sur la machine compromise, le malware interroge un LLM hébergé en Chine pour décider de ses prochaines actions. Très concrètement, le logiciel malveillant s’adapterait à l’environnement qu’il découvre et rendrait ainsi les signatures de détection classiques largement inopérantes.

  • Deuxième cas, côté chinois cette fois : une campagne de cyber-espionnage dont l’essentiel des décisions tactiques (choix des cibles, mouvements latéraux, exfiltration) a été délégué à un système d’IA. “L’opération était majoritairement orchestrée par IA, avec une intervention humaine ponctuelle”, résume Sandra Joyce. C’est, à sa connaissance, la première campagne de cette nature documentée publiquement.

  • Le signal le plus inquiétant est, selon Sandra Joyce, l’effondrement de la barrière d’entrée des nouveaux arrivants.

  • Plus étonnant, les LLM permettent également aux groupes de pirates spécialistes du ransomware de cibler de nouveaux marchés. Historiquement, les groupes ransomware ciblaient massivement les pays anglophones, faute de moyens pour produire à l’échelle des messages de phishing crédibles dans d’autres langues. “Nous avons vu une augmentation très importante”, indique-t-elle, en pointant l’Allemagne du doigt, qui a même surpassé le Royaume-Un

• Is AI putting graduates out of work already? (The Economist)

  • We found that graduates in fields more exposed to AI have suffered markedly worse outcomes.

  • Between 2022 and 2024 graduates in the least-exposed 20%—studying subjects such as education, philosophy and civil engineering—saw their average full-time employment rate fall by just 1.5% points.

  • Those in the most exposed 20%—including computer science, computer engineering and information science—suffered a 6.6%-point drop.

  • We updated these figures for the most exposed fields, using data from 13 universities, and found that the trend continued for the class of 2025. The rate of full-time employment fell from nearly 70% to 55% in three years—notably, the 3 years following ChatGPT’s release in 2022. Prior to that, it had been stable.

  • Students are already changing course.

  • Enrolment in computer programming, which focuses on coding skills rather than theory, dropped by 26%.

• This driverless Chinese mining truck is giant, agile, and shows the industrial future of AI (source)

  

  • Equipped with a Level 4 driving brain that allows it to operate with no human intervention, this massive robot on four wheels can literally move on a dime, rotating 360 degrees on its own vertical axis and moving sideways like a crab, operating 24/7.

  • “The Shuanglin K7 measures 14 meters in length and 5.70 meters in width, weighing 100 tons when empty.”

    • For comparison :

      • An empty semi-truck (tractor and trailer) typically weighs between 15 and 17.5 tons in the United States

      • The heaviest truck in the world currently in operation is the BelAZ 75710, a colossal mining dump truck. When completely empty, it weighs 360 tons and measures 20.6 m in length, 10 m in width, and 8.2 m in height.

  • The Chinese government—which depends on minerals to keep its control of the global supply chain—needs this vehicle to fully automate Chinese extraction by 2030, a goal already materializing in Xinjiang and Inner Mongolia with other less capable fleets.

• Parlons passé : Neanderthals went to the dentist (really) (The Economist)

  • A 59,000-year-old molar from an adult Neanderthal has a large hollow on the chewing surface that reaches deep into its centre, where nerves and blood vessels would have been.

  • This suggests that a Neanderthal had intentionally drilled into the tooth to scoop out infected tissue.

  • This dental discovery adds valuable detail to the picture of Neanderthal life. For one thing, say the authors, it shows they must have had fine motor control. More important, it suggests that they would have understood the importance of enduring some immediate (and excruciating) pain to reduce the chance of death.

  • Such a calculation would have been impossible without causal reasoning, an attribute whose presence among Neanderthals has long been debated.

  • This finding helps to bring that debate to a close.

• Parlons passé : how disruptive was the Industrial Revolution for employment and wages? (The Economist)

  • GDP per person in America, which had by then dislodged Britain as the world’s frontier economy, grew by 2.5% a year, the fastest ever for a leading economic power.

  • The level of job disruption, as measured by the share of employment shifting between industries or occupations, was at times more than twice as high as it is today.

  • Yet many people look back fondly on that era as a time of rising wages, widening opportunity and unpolarised politics.

  • Actually, the average employer paid workers reasonably fairly after selling his wares and deducting the cost of materials. He did not profit from exploiting his staff, as Engels supposed.

  • The problem for labourers was less unfair pay than sharp rises in the cost of living. Food prices rose steadily, and sometimes soared, because of war and high tariffs on grain imports. The villains of the Industrial Revolution were politicians, not machines.

---

The Internet Can’t Stop Watching Figure AI’s Humanoid Robots Handling Packages (ArsTechnica, the 12-hour video lol)

• “The promotional robot demo has become a viral sensation among tech enthusiasts, spurring YouTube commenters to name the robots and the company to rapidly roll out related robot merchandise in response. ...But despite such sentiments, it’s worth bearing in mind that even the most impressive robot demos represent narrow windows for understanding real-world robot capabilities.”

• Users on X described the livestream in glowing terms, such as “the greatest product demo since Steve Jobs’ ‘one more thing.’”

• Figure’s event began on May 13 as a planned 8-hour robot demonstration featuring the company’s latest Figure 03 robots. The chosen robotic task involved inspecting the bar codes on various small packages—including cardboard boxes and soft padded envelopes or bags—and then placing the packages on a conveyor belt with the bar codes facing downward.

• On May 17, Adcock laid out the “Man vs. Machine” scenario for the head-to-head competition that would last 10 hours.

• A comparison video revealed some of the current differences between the humanoid robots’ performance and human capabilities. Whereas Gérard could speedily and precisely pick up packages for inspection before moving them along, the robots moved at a slower but methodical pace. The robots also sometimes spent extra time trying and failing to pick up packages or grabbed repeatedly at empty air when trying to sweep packages along.

• That difference allowed Gérard to stay ahead on the package task until he took a mandated break, according to Business Insider. But he quickly recovered the human lead from the robotic competition once he returned—eventually claiming victory by sorting 12,924 packages versus the robots’ 12,732 packages.

• Figure’s website describes the robots’ whole-body controller system as having been trained on more than 1,000 hours of human motion data, along with spending time training in simulation across more than 200,000 parallel environments.

• My take: as Yann LeCun would say, it’s a rather repetitive task, and the robot had to be trained on more than 1,000 hours of video for something a human could learn in seconds, it’s great, may help us do more work more cheaply and free up human time for more productive endeavours, BUT we are still very far from general-purpose robots that could enter a room and complete a new mission autonomously.

---

📢 👽Alerte : update sur le paradoxe de Fermi: where are the aliens??? 👾🛸(Par Casey Handmer, résumant les travaux de Robin Hanson)

• 3 grandes classes d’explication :

• 1.Interstellar travel might be impossible.

  • Antimatter might be energetic enough to accelerate to close to light speed, but that doesn’t mean that interstellar travel is possible – colliding with a single dust grain would be very bad news. Perhaps the galaxy has a million technological civilizations, and they’re all trapped in their respective solar systems by the enormous gulfs of space.

• 2.The Great Filter.

  • Maybe intelligent alien life is rare because there’s some filter or set of filters that kills off life forms that get too advanced. This filter could be in our past (multicellularity, asteroid extinction, solar flares) or in our future (nuclear war, hostile aliens killing upstarts, AIs starving us to death, depopulation, loss of culture of exploration). But you only need one very powerful alien species to overcome these filters and then they can fill up the galaxy relatively quickly. As far as we can see, the galaxy is not full.

• 3.Near light speed travel is hard to observe for people at the destination.

  • If aliens were expanding at a high fraction of the speed of light (c), the light carrying information of their coming would be only just ahead of them. So even though aliens might be quite close, we wouldn’t see them until just before they arrived.

  • The universe could be in three different states, observationally.

    • a. What we observe (no aliens),

    • b. aliens seen but not here yet,

    • c. and aliens among us.

  • But if the aliens we see are traveling at high speed toward us, the intermediate state (seen but not met) is unlikely to be longer than a handful of weeks.

  • Choosing our present time at random, there is almost zero chance for humanity to find itself in a time where we’re aware of alien intelligences but haven’t yet met them. That is, Earth is 4.5 billion years old (no aliens), then one day the Vera Rubin Observatory sees a flash that turns out to be an alien spacecraft departing to meet us from 100 light years away, traveling at 99.9% of c (99.9% de la vitesse de la lumière). They arrive just five weeks later. For the remaining billions of years of Earth’s existence, we are in the world of aliens among us.

  • I think it’s physically possible to reach 99% of c with current human technology, so there’s no reason to suppose aliens with better technology would fly slower than this, and they could fly much faster.

  • Another example: if they’re traveling at 99% c, we will see them (in the sense we’ll see the flash of their departure) only when they’re 99% of the way here. If they’ve traveled 1000 light years to visit us, we’ll see them (at best) 10 years before they arrive. We might not see them depart at all – 1000 light years is far enough away that some stars are too dim to see with the naked eye. Meanwhile, 1000 light years is a long way to go, so it’s fortunate that at high speed, relativistic time dilation kicks in and helps to pass the time. At 99% c, the 1000 light year trip only feels like 142 years.

  • But if they can hit 0.99c, why not 0.999c?

  • At 99.9% c, the trip will feel like only 45 years to them, and we will get a whole year of warning, assuming we see them launch 1000 light years away.

  • Another example, let’s say they’re doing 99.5% c over 500 light years. They will be in flight for 502.5 years, they will arrive 2.5 years after their light reached us, and on board they’ll feel just 50 years pass by.

Partagez cette newsletter par Whatsapp en cliquant ici ❤️

• Si vous appréciez cette newsletter, 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

👋👀🦉💡 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” 👇

• 🔒 Why AIs may not need superhuman reasoning abilities in order to achieve superintelligence

• 🔒 OpenAI has the best model, so why is Anthropic winning? Here is why

• 🔒 Why the threat from new pathogens designed with AI help is an much graver danger than AI-backed hackers: what we can and cannot do about it

• 🔒 Pourquoi plus la superintelligence pourrait devenir une sacrée glandeuse

Les précédentes newsletters

❤️ Transférez cette newsletter à vos amis, ou partagez-la par Whatsapp en cliquant ici.

Pour vous inscrire et ne pas manquer les prochaines éditions, vous pouvez cliquer ici.

Merci, et à bientôt !

Thomas