🚀 Best videos done with best-in-class Google AI tool; Rats Learned to Drive ; one-person unicorn & more
Bonjour,
Vous recevez 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.
Et voici donc ma toute dernière sélection !
L’apéro
Les derniers épisodes du podcast Parlons Futur (Spotify, Apple, Youtube), notamment :
Dernière vidéo du robot humanoïde de Boston Dynamics : Tesla ringardisé, quoi de neuf ? What's next?
Robots humanoïdes, robots chiens : changeront-ils l'art de la guerre significativement ?
L'IA peut-elle créer de faux souvenirs ? Retours et échanges suite à une nouvelle étude
L'ultra-personnalisation des services permise par la data et l'IA est-elle souhaitable ?
A twitter thread with 10 amazing short robot videos of all shapes and forms!
L'économiste Olivier Blanchard dit : "Amazing. Hard to think it is not going to show up in productivity numbers."
Elon Musk’s humanoid robot catches tennis balls with new hand upgrade: see the 15-sec video
Meet 'Blackbird': A flying taxi concept by an Austrian company that spins and moves in any direction thanks to new propulsion system
...but many now think that flying cars are a dead-end : "Recently there’s been enough news of very predictable — and predicted — failures" says tech analyst Michael Barnard about Electric vertical take-off and landing aircraft (EVTOLs)
Kitty Hawk Corporation, the Silicon Valley-based EVTOL pioneer backed by Google co-founder Larry Page, announced in September 2022 that it was ceasing operations.
Rolls-Royce’s ambitions in the electric flying taxi space came to an end in September of 2024, as the British aerospace giant shut down its Advanced Air Mobility division.
Vertical Aerospace, a prominent UK-based EVTOL company, entered advanced negotiations with creditors in November 2024
Joby continues to pretend it’s going to have aircraft certified and in the air real soon now. The company originally pretended that its first-of-a-kind aircraft with multiple novelties and a class history of falling out of the sky and killing people would be certified and available for sale in 2023. Now they are pretending it’s 2025.
Volocopter, a German sibling firm of already failed Lilium, is struggling as well.
Commonwealth Bank of Australia says GenAI helped them lower frauds by 30%. (Bloomberg)
Ce nouveau radar boosté à l’IA permet de détecter les automobilistes trop alcoolisés (Figaro)
La caméra du radar «surveille de nombreux paramètres, par exemple la façon dont une voiture se déplace sur la route»
OpenAI's Sam Altman very bullish on Artificial General Intelligence and Superintelligence (from his blog)
We are now confident we know how to build AGI as we have traditionally understood it. We believe that, in 2025, we may see the first AI agents “join the workforce” and materially change the output of companies. We continue to believe that iteratively putting great tools in the hands of people leads to great, broadly-distributed outcomes.
We are beginning to turn our aim beyond that, to superintelligence in the true sense of the word. We love our current products, but we are here for the glorious future. With superintelligence, we can do anything else. Superintelligent tools could massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own, and in turn massively increase abundance and prosperity.
This sounds like science fiction right now, and somewhat crazy to even talk about it. That’s alright—we’ve been there before and we’re OK with being there again. We’re pretty confident that in the next few years, everyone will see what we see, and that the need to act with great care, while still maximizing broad benefit and empowerment, is so important. Given the possibilities of our work, OpenAI cannot be a normal company.
Pour contexte:
AGI, or artificial general intelligence, is a nebulous term, but OpenAI has its own definition: “highly autonomous systems that outperform humans at most economically valuable work.”
Interestingly, here's Microsoft own definition of AGI: AI systems that can generate at least $100 billion in profits. (and coincidentally, when OpenAI achieves this, Microsoft will lose access to its technology, per an agreement between the two companies.)
Lol: OpenAI’s Super-Hyped Sora Goes Absolutely Freakshow If You Ask It to Generate Gymnastics Videos : see the short videos on X
But at least AI is now better at generating videos of Will Smith eating pasta: see the 20-sec video showing one year ago or so and now (not sure what tech is used now, likely Veo 2, see below)
Google published a new video generator, Veo 2, which is far ahead of OpenAI’s Sora and arguably the best available right now.
Tech Analyst Ben Evans says: "It’s easy to watch the demos once and be amazed, and then twice and notice the issues… but you also have to think about how much of that will get fixed in a year, or can be fixed in post, how much footage comes back from a day’s shooting that also has to be fixed, and to wonder what it means that anyone can have tools like this."
A creator made a movie scene out of it, to be watched here, he detailed:
Introducing, The Heist. Every shot of this film was done via text-to video with Veo 2. Thousands of generations to get the final film, but I am absolutely blown away by the quality, the consistency, and adherence to the original prompt. When I described “gritty NYC in the 80s” it delivered in spades - CONSISTENTLY. While this is still not perfect, it is, hands down, the best video generation model out there, by a long shot. Additionally, it’s important to add that no VFX, no clean up, no color correction has been added. Everything is straight out of Veo 2. Google DeepMind
Comparing 8 video generation AI tools with prompt "slicing steak", including Veo 2: see the 1-min video, it's very clear who's the winner, by far!
Another comparison Sora vs.Veo-2. (feels like comparing a bike vs. a starship)
More examples of videos done with Veo 2
"1960s film of Elvis shaking hands with an alien in the White House"
"A cat roars while looking at its reflection in the mirror but instead sees itself as a lion roaring"
Seems to handle "physics" rather well, was put to a test by a user with a range of outputs to see how liquids and solids would react with each other. (17-sec video)
Meta's updates
Nvidia’s Jensen Huang recently spoke about the role agents already play at Nvidia and how he sees their future (source):
Brad: Are you already using chain-of-reasoning and tools like o1 in our own business to improve it?
Jensen: Absolutely. Our cybersecurity system today can’t run without our own agents. We have AI agents helping design chips—Hopper wouldn’t be possible, Blackwell wouldn’t be possible and don’t even think about Rubin. We have AI chip designers, AI software engineers and AI verification engineers, and we build them all internally. We have the ability, and we’d rather use the opportunity to explore the technology ourselves. I’m hoping that Nvidia someday will be a 50,000-employee company with 100 million AI assistants […] AIs will recruit other AIs to solve problems […] and so we’ll just be one large employee base, some of them digital and some biological.
Sam Altman anticipates the emergence of a one-person unicorn—a billion-dollar enterprise managed by a single individual leveraging an army of AI agents. (source)
DeepMind unveiled an AI that predicts weather 15 days in advance in minutes rather than the hours, if not days, usually needed with traditional models. (source)
In a head-to-head with the European Center for Medium-Range Weather Forecasts’ model (ENS)—the best “medium-range” weather forecaster today—the AI won over 90 percent of the time
Google toujours : "Wow the NotebookLM update blows my mind You can now participate in real time in the AI-generated podcast from your own sources " (source)
Chinese's Ai model DeepSeek outwits the top American labs despite sanctions! (by Azeem Azhar)
DeepSeek, a Chinese AI company that has spun out of a hedge fund, released its Version 3 large language model. It performs about as well as Claude 3.5 Sonnet. The breakthrough is how cheap it was to train. At roughly $5.6m to train, DeepSeek v3 was about a hundred times cheaper than Meta’s Llama 3 and probably a similar magnitude cheaper than Claude.
DeepSeek’s API costs 12 times less than Anthropics. Developers around the world are starting to have fun with it.
Biden’s chip export controls have notionally hampered the team. But necessity is the mother of invention, and they delivered this superb model through ingenuity rather than brute force.
The ramifications:
As was reasonably likely, the export controls have backfired.
The gap between open source (like DeepSeek) and closed source (like OpenAI) is narrowing rapidly. It calls into question efforts to constrain open-source AI development.
China now has several world-class teams building AI systems, including Qwen, Kling (in video), Apollo (in self-driving) and, of course, Bytedance.
Doctors Intrigued by Treatment That Makes Dead Brains Show Signs of Life (source)
Scientists were astonished to find that recirculating a cocktail of preserving agents through a severed pig's head caused the animal's brain to show signs of life.
As New Scientist reports, basic cellular functions were restored in the dismembered brain — something that was previously thought impossible following the cessation of blood flow.
"We had to develop new methods to make sure no electrical activity is occurring in an organized way that might reflect any kind of consciousness," Vrselja told New Scientist.
In a 2019 experiment involving pig brains, the researchers managed to bring some activity back four hours after decapitation.
For now, they're using their invention to test out treatments for Alzheimer's and Parkinson's.
Similar techniques could also be used to prolong the shelf life of donor organs, which could save lives.
These Rats Learned to Drive—and They Love It (Wired)
We crafted our first rodent car from a plastic cereal container. After trial and error, my colleagues and I found that rats could learn to drive forward by grasping a small wire that acted like a gas pedal. Before long, they were steering with surprising precision to reach a Froot Loop treat.
While we can’t directly ask rats whether they like to drive, we devised a behavioral test to assess their motivation to drive. This time, instead of only giving rats the option of driving to the reward, they could also make a shorter journey on foot – or paw, in this case.
Surprisingly, two of the three rats chose to take the less efficient path of turning away from the reward and running to the car to drive to their destination. This response suggests that the rats enjoy both the journey and the rewarding destination.
Incroyable :
Neuroscientist Curt Richter also made the case for rats having hope. In a study that wouldn’t be permitted today, rats swam in glass cylinders filled with water, eventually drowning from exhaustion if they weren’t rescued.
Lab rats frequently handled by humans swam for hours to days.
Wild rats gave up after just a few minutes. If the wild rats were briefly rescued, however, their survival time extended dramatically, sometimes by days.
It seemed that being rescued gave the rats hope and spurred them on.
Cops Say Insurance CEO Shooter's Pistol and Silencer Were Both 3D-Printed (source)
mm.. there seems to be a Stupidly Easy Hack Can Jailbreak Even the Most Advanced AI Chatbots, according to research from Anthropic itself (source)
Along with capitalization changes, prompts that included misspellings, broken grammar, and other keyboard carnage were enough to fool these AIs — and far too frequently.
The work illustrates the difficulties of "aligning" AI chatbots, or keeping them in line with human values
Fun fact: Brazil's northernmost point is closer to every country in the Americas than to Brazil's southernmost point (source with map)
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
Waymo still doing better than humans at preventing injuries and property damage (Verge)
They found that the performance of Waymo’s vehicles was safer than that of humans, with an 88 % reduction in property damage claims and a 92% reduction in bodily injury claims.
Across 25.3 million miles, Waymo was involved in:
9 property damage claims (the average human driving a similar distance would be expected to have 78 property damage)
2 bodily injury claims (against 26 bodily injury claims for an average human driving a similar distance)
Waymo’s vehicles also performed better when compared to new vehicles equipped with all the latest safety tech, including automatic emergency braking, lane-keep assist, and blind spot detection. When compared to this group, Waymo’s autonomous driving system showed an 86 percent reduction in property damage claims and a 90 percent reduction in bodily injury claims.
About dark factories in China (Thomas Friedman, NYT)
A retired Chinese official mentioned to me in passing over dinner that she wanted to buy a new high-tech bed and decided to go see the offerings at the factory. When she arrived, though, she found it was a “dark factory” — so the lights were turned on just for her. It wasn’t dark because it was out of business, she told me. It was dark because it was so fully roboticized that the company doesn’t waste electricity keeping the lights on for any humans — except for the engineers who come to clean or adjust the machines once a day.
As an article in the state-run China Daily explained: “From steel plates and mobile phones to household motors and rocket ignition device parts, more business lines in China are using artificial intelligence to power their production and have introduced ‘dark factories’ with their 24-hour uninterrupted and unattended production capabilities. Dark factories, also called smart factories, are entirely run by programmed robots with no need for lighting.”
Noah Smith on the scale of Chinese manufacturing capabilities (as quoted by Thomas Friedman, NYT)
In 2000, “the United States and its allies in Asia, Europe and Latin America accounted for the overwhelming majority of global industrial production, with China at just 6 percent even after two decades of rapid growth.”
By 2030, Smith wrote, the U.N. agency predicts “China will account for 45 percent of all global manufacturing, single-handedly matching or outmatching the U.S. and all of its allies.
“This is a level of manufacturing dominance by a single country seen only twice before in world history — by the U.K. at the start of the Industrial Revolution, and by the U.S. just after World War II.” Smith wrote, “It means that in an extended war of production, there is no guarantee that the entire world united could defeat China alone.”
Je vous invite maintenant à vous abonner à la version payante pour recevoir la semaine prochaine les résumés de ces longs articles le plus souvent issus des plus grands journaux (et accéder aux archives) en cliquant sur le bouton “Upgrade to paid/Subscribe now” plus bas :
How good is Google’s “Deep Research” new AI tool ?
How building Telescopes on the Moon Could Transform Astronomy
How antimatter engines could revolutionize space travel
How Lab-grown models of embryos increasingly resemble the real thing
Les dernières newsletters :
L’addition ?
Cette newsletter est gratuite, si vous souhaitez m'encourager à continuer ce modeste travail de curation et de synthèse, vous pouvez prendre quelques secondes pour :
transférer cet email à un(e) ami(e) ou partager par whatsapp
et ici pour la version 100% en anglais
étoiler cet email dans votre boîte mail
cliquer sur le coeur en bas d’email
Un grand merci d'avance ! 🙏
Ici pour s’inscrire et recevoir les prochains emails si on vous a transféré celui-ci.
Merci, et à bientôt !
Thomas