🚀World’s largest aircraft; le SpaceX français ?; crazy speech-to-speech, text-to-video, text-to-music AIs demos & plus

Fastest sail boats ; chenille fabrique une tente (!) & more

Dec 05, 2023

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, plus d'infos sur moi en bas d'email.

Voici donc ma dernière sélection !

Un long apéro this week…

Google has quietly postponed the launch of Gemini—its much-awaited new AI model
- Google is struggling to match OpenAI's capabilities in conversational AI, causing it to delay Gemini's launch to continue improvements.
- Catching up with Gpt-4 is very hard. It’s been 9 months since its launch and all the newer models are still way behind.
Wow: Generating one image takes as much energy as fully charging your smartphone, according to the study from researchers at the AI startup Hugging Face and Carnegie Mellon University.
Cool new AI service: Pika, the idea-to-video tool: create or edit videos using words: see the 1-min demo
And another very cool AI tool by Meta: speech-to-speech translation that preserves the voice, the tone, and the expression. (listen to the 15-sec example from 1:00 to 1:15)
Meta’s AI chief Yann LeCun thinks current AI systems are decades away from reaching some semblance of sentience, and is skeptical on quantum computing (source)
- “Text is a very poor source of information,” “There’s a lot of really basic things about the world that they just don’t get through this kind of training,”
- Unlike Google, Microsoft and other tech giants, Meta is not making a big bet on quantum computing.
- “The number of problems you can solve with quantum computing, you can solve way more efficiently with classical computers,” LeCun said.
- “Quantum computing is a fascinating scientific topic,” LeCun said. It’s less clear about the “practical relevance and the possibility of actually fabricating quantum computers that are actually useful.”
Lyria, by Google DeepMind: the world's most sophisticated AI music generation system (source)
- Qq exemples impressionnants !
- Produire une instrumentale juste en fredonnant : 45-sec video
- Transforming beatboxing into a drum loop: 20-sec video
- Transforming singing into an orchestral score: 20-sec video (wow)
- Transforming chords from a MIDI keyboard into a realistic vocal choir: 40-sec video
- Lyria can also produce compelling music & vocals from just a text prompt
Amazing demo : cette IA vous analyse et génère en temps réel des commentaires sur votre activité avec la voix du célèbre narrateur de documentaires animaliers David Attenborough : watch the 1-min video!
- "We observe the sophisticated Homo sapiens engaging in the ritual of hydration."
Using AI to help read medical exams led to the finding of 13% more breast cancers. Of 30 extra cancers detected, 83% were invasive types that would’ve been missed without AI. (source)
Bonnes trouvailles de la newsletter TTSO sur l'IA:
- Les visages générés par l’IA ne sont pas seulement crédibles, ils sont jugés (par les humains) plus humains que les vrais visages humains. (source)
- On demande à ChatGPT de répondre à 50 questions de rubriques "courrier du cœur". Les 404 personnes interrogées trouvent ces réponses "plus équilibrées, plus complètes, plus empathiques, plus utiles, et meilleures" que celle de l’expert(e) humain(e). (source)
- Les réponses d’un chatbot à des questions médicales posées sur un forum sont meilleures que celles des médecins. Elles sont surtout bien plus empathiques (45% des réponses de l’IA le sont, contre 5% de celles des médecins). Détail important : ce sont des professionnels de santé qui en jugent ! (source)
- How do you see your orgasms? Here's what they look like according to artificial intelligence
Musk says a new Starship will be ready to fly again before the end of the year !!
- Successes from last flight in November :
- All 33 Raptor engines worked. (At least six hadn’t worked properly during the April test flight.)
- Starship then successfully separated from the Super Heavy booster (see picture above). “This was the first time this technique has been done successfully with a vehicle of this size,” SpaceX officials wrote in their post.
- The vehicle reached an altitude of about 150 kilometers, “becoming the first Starship to reach outer space and nearly completing its full-duration burn.”
- “Just inspected the Starship launch pad and it is in great condition! No refurbishment needed to the water-cooled steel plate for next launch,” Musk wrote
- ça s’accélère : The time between the first and second Starship flights took seven months. Next one possibly before the end of 2023!
Wow: 40% of a top scientist’s time is spent on things other than research, such as looking for money (The Economist)
Xavier Niel dévoile Kyutai, son nouveau laboratoire d'intelligence artificielle
- Annoncé le 26 septembre dernier, ce dernier à vocation à contribuer à la démocratisation et au développement de l'intelligence artificielle générale, une IA dotée de capacités similaires voire meilleures à celles du cerveau humain. Pour y parvenir, le laboratoire bénéficie d'une enveloppe de 100 millions d'euros apportée par Xavier Niel, auxquels s'ajoutent 200 millions apportés par Rodolphe Saadé, PDG de CMA CGM et Eric Schmidt, ex CEO de Google, soit un total de 300 millions
The best explanation of what happened at OpenAI (Ben Evans): the people who think we should slow down and be careful mounted a coup against the people who think we should speed up and be careful.
- Note that Sutskever moved to work on ‘alignment’ (trying to work out how to make an AGI be nice to us) this summer.
- The problem, and the conflict at OpenAI, is that no one knows about the AI risk. We don’t have a clear and accepted theoretical model for what intelligence is, nor how we would plot our own intelligence within that as compared to, say, dogs, nor where LLMs are in relation to that, nor how quickly they might converge. We don’t know. And so every conversation descends to argument by analogy, or from authority, or to a long list of other logical fallacies. Yes, Geoff Hinton is worried, but he’s worried because he doesn’t know, not because he does, which is an odd kind of argument from authority. No-one knows.
- “The board can fire me, I think that’s important,” Altman told Bloomberg in June.
  - “It turns out that they couldn’t fire him, and that was bad,” says Toby Ord, senior research fellow in philosophy at Oxford University, and a prominent voice among people who warn AI could pose an existential risk to humanity. (Wired)
- Also: OpenAI researchers warned board of AI breakthrough ahead of CEO ouster, sources say (Reuters)
Amazing Nissan LEAF1-min video ad "Gas Powered Everything" - I'm not sponsored ;)
When Bell Labs announced the transistor – arguably the most important invention of the modern age – the New York Times buried the story on page 46.
Bill Gates : "AI Agents are not only going to change how everyone interacts with computers. They’re also going to upend the software industry, bringing about the biggest revolution in computing since we went from typing commands to tapping on icons." (source)
"Science is the only news." (Whole Earth Discipline, by Stewart Brand)
- " When you scan through a newspaper or magazine, all the human interest stuff is the same old he-said-she-said, the politics and economics the same sorry cyclic dramas, the fashions a pathetic illusion of newness, and even the technology is predictable if you know the science."
- "Human nature doesn’t change much; science does, and the change accrues, altering the world irreversibly."
Spinal Implant Helps a Man With Severe Parkinson’s Walk With Ease Again (source)
- To personalize his implant, the team captured hours of video of his walking patterns. They then built a model of his muscles, skeleton, and several joints, such as the hips and knees. Using the model, they trained software to compensate for any dysfunction—allowing it to decipher the user’s intent and translate it into electrical zaps in the spinal cord to support the movement.
Noise-canceling headphones could let you pick and choose the sounds you want to hear (MIT Tech Review)
- Future versions of the technology could let users opt back in to certain sounds they’d like to hear, such as babies crying, birds tweeting, or alarms ringing.
Cool demo : Two 19-year-old students have developed a pair of gloves that convert sign language to speech or text. (50-sec video)
8-min Short-movie depicting a nuclear escalation triggered by AI (Yann LeCun would find it ridiculous)
- This work of fiction seeks to depict key drivers that could result in a global Al catastrophe:
  - Accidental conflict escalation at machine speeds;
  - Al integrated too deeply into high-stakes functions;
  - Humans giving away too much control to Al;
  - Humans unable to tell what is real and what is fake, and;
  - An arms race that ultimately has only losers.
- The sequel to this video: How would a nuclear war between Russia and the US affect you personally? (4-min video)
The First Small-Scale Nuclear Plant in the US Died Before It Could Live (Wired)
- The utilities backing the plant were spooked late last year by a 50% increase in the projected costs for the project—even after factoring in substantial funds from the Inflation Reduction Act.
- “One of the stories they’ve kept telling people was that the SMR was going to be a lot cheaper than large-scale nuclear,” David Schlissel, an analyst at the nonprofit Institute for Energy Economics and Fiscal Analysis, told WIRED last month. “It isn’t true.”
Original : un balcon qui se replie pour se transformer en fenêtre (What the f...) (15-sec video)
Une nouvelle mouvance dans la Silicon Valley : Effective accelerationism, or E/acc (NYT)
- “Effective accelerationism aims to follow the ‘will of the universe’: leaning into the thermodynamic bias towards futures with greater and smarter civilizations that are more effective at finding/extracting free energy from the universe,” and “E/acc has no particular allegiance to the biological substrate for intelligence and life, in contrast to transhumanism.”
How mining could be disrupted before 2050 (Casey Handmer, founder of Terraform Industries, whom I interviewed here about his revolutionary startup and more)
- There’s no reason that our pursuit of valuable minerals in the crust needs to start by destroying large swaths of the upper few feet that are so critical to life. Instead, we will see the commercialization of integrated tunnel boring machines of enormous size that will extract valuable minerals at depth before extruding a narrow, fast-moving wire of any desired alloy through a small, surface-mounted “well”. Ordinarily, these machines will back-fill their tunnels with the dross (débris) but could also carve out enormous underground caverns for any desired use, be it high speed trains, vertical farming, nuclear reactors, underground rivers, or secret realms.
By 2050 it may be the case that solar panels and chemistry give us the ability to make most of our food with 0.1% of the land we currently use for farming (Casey Handmer)
- Early studies today are finding ways to electrically synthesize starches, proteins, and fats from water and air. By 2050 it may be the case that solar panels and chemistry give us the ability to make most of our food with 0.1% of the land, reducing the enormous and catastrophic environmental impact of our civilization farming literally 40% of the Earth’s entire land surface area, while reducing food costs and improving food security.
Capacités animales incroyables : cette chenille se fabrique une tente , sans rigoler (1-min video)
Mind-blowing drone show in China (1-min video)
Sensations fortes : 2 équipes francophones vont tenter de battre le record de vitesse avec un "bateau" à voile (The Economist)
- To qualify, a craft has to be able to float, have at least one person on board, be propelled only by the wind and be in contact with the water
- Le bateau à voile qui détient le record de vitesse : 65 knots, établis en 2012, encore invaincu, 120km/h
- Le bateau de la première équipe en lice
- Le bateau de l'autre équipe
- Drôles de bateaux...
ReelShort is the latest Chinese export to conquer America (The Economist)
- American viewers can tune in to dozens of similar rapid-fire dramas, with titles such as “The Double Life of My Billionaire Husband” and “Son-in-Law’s Revenge”, on an app called ReelShort.
- If it all seems a bit foreign, then that is because it is. ReelShort is owned by COL Group, a digital publisher based in Beijing. Some of its shows are adapted by Chinese teams at COL’s Californian subsidiary, Crazy Maple Studio, from Chinese scripts that were first written and produced for audiences in China.
Prompt Pour créer un logo professionnel à partir d'un croquis avec GPT-4 Vision et Dall-E 3 (JDN)
- “En tant que graphiste avec une expérience de 10 ans, tu es habitué à transformer des idées abstraites en designs visuels captivants. Tu viens de recevoir un croquis de logo en pièce jointe, et ta tâche est de le reproduire fidèlement en utilisant Dall-E3. Concentre-toi sur l'harmonie des couleurs, en veillant à ce que le contraste soit suffisant pour une clarté optimale. Assure-toi que les lignes et les formes soient équilibrées et symétriques, reflétant un sens aigu de la proportion et de l'esthétique. Le logo doit être adaptable, facilement reconnaissable et mémorable, tout en conservant une simplicité élégante. Utilise ton expertise pour interpréter le croquis de manière créative, en ajoutant ta touche personnelle tout en restant fidèle à la vision originale.”

Partagez cette newsletter par Whatsapp en cliquant ici ❤️

Si vous appréciez cette synthèse gratuite, n’hésitez pas à prendre 3 secondes svp pour l’envoyer à ne serait-ce qu’un contact par Whatsapp en cliquant ici 🙂
- Partagez la version 100% en anglais de la newsletter avec vos amis non-francophones en cliquant ici
Et si on vous a partagé cet email, vous pouvez cliquer ici pour vous inscrire et ne pas manquer les prochains
Pourriez-vous me dire en retour d’email si vous seriez plus enclin à partager cette newsletter si elle était traduite entièrement en français ?

À table !

Latitude, le futur SpaceX français ? (source)

Fondée en 2019 et aujourd’hui basée à Reims, la start-up progresse très efficacement vers le premier vol de son micro-lanceur nommé Zéphyr. Ce 21 novembre, Latitude change de phase dans sa jeune vie, passant du stade de la Recherche & Développement à l’industrialisation.
Latitude est désormais la seule compagnie française avec ArianeGroup à passer le cap de la production industrielle.
Zéphyr est cependant bien plus petite qu’Ariane 6, mesurant seulement 17 mètres de haut. Elle est capable de livrer jusqu’à 200 kilos de satellite en orbite basse. Elle est donc destinée à prendre en charge des microsatellites et des nanosatellites, toujours plus nombreux à être utilisés de nos jours.
L’extension de l’usine permettra de lancer la production du micro-lanceur. Une autre usine ouvrira plus tard à Champigny-sur-Marne, elle permettra à terme de produire 50 Zéphyr par an, en attendant que Latitude nous propose des lanceurs plus gros.
À cette extension, s’ajoutent quatre hectares fraîchement acquis à l’aéroport de Vatry, pour tester les moteurs, seuls mais aussi les étages intégrés de la Zéphyr. « Ce sera le second centre d’essai de ce genre en France métropolitaine avec le centre historique de Vernon », se réjouit Stanislas Maximin, ce dernier étant piloté par ArianeGroup, considéré par un concurrent direct par Latitude.
50 nouveaux employés depuis 2021, et 40 supplémentaires pour atteindre un total de 120 employés en janvier ! Le tout est réussi avec seulement 20 millions d’euros provenant de fonds privés et publics. En comparaison, des concurrents allemands ont levé plus de 100 millions d’euros pour arriver à un stade à peine plus avancé.

Effect of AI on the music industry (The Economist)

Spotify now adds more than 100,000 new tracks every day, many of them AI-made.
In 2017 artists who are signed to record labels accounted for 87% of the streams on Spotify. Last year their share was only 75%.
Amid an explosion in online content on platforms from YouTube to TikTok, fans have flocked as never before to the biggest acts. It has been a good time to be an amateur creator, but an even better time to be a superstar.
At the same time, data from Spotify show that between 2017 and 2022, as the platform was flooded with tens of millions of amateur tracks, the number of artists making at least $1,000 a year in royalties increased by 155%, the number making $5m or more increased by 165%, and the handful of headliners making $10m or more increased by 425%.
Those who have done least well are middling-to-big artists, who face more competition from entertainment’s long tail but have been unable to break into the elite group at the top.

Google DeepMind’s new AI tool helped create more than 700 new materials (MIT Tech Review)

From EV batteries to solar cells to microchips, new materials can supercharge technological breakthroughs.
But discovering them usually takes months or even years of trial-and-error research.
A new tool from Google DeepMind uses deep learning to dramatically speed up the process of discovering new materials.
The technology has already been used to predict structures for 2.2 million new materials and 380,000 new stable materials—many counter to human intuition, of which more than 700 have gone on to be created in the lab and are now being tested.
Thanks to that tool named GNoME, the total number of stable materials known to us has grown almost tenfold, to 421,000
Berkeley Lab’s new autonomous laboratory, named the A-Lab, has been using GNoME’s discoveries, integrating robotics with machine learning to optimize the development of such materials.
The lab is capable of making its own decisions about how to make a proposed material and creates up to five initial formulations. These formulations are generated by a machine-learning model trained on existing scientific literature. After each experiment, the lab uses the results to adjust the recipes.
Researchers at Berkeley Lab say that A-Lab was able to perform 355 experiments over 17 days and successfully synthesized 41 out of 58 proposed compounds. This works out to two successful syntheses a day.
In a typical, human-led lab, it takes much longer to make materials. “If you’re unlucky, it can take months or even years,”
Researchers at DeepMind and Berkeley Lab say these new AI tools can help accelerate hardware innovation in energy, computing, and many other sectors.
Together, the collaboration could launch a new era of materials science. “It’s very impressive,” said Dr. Andrew Rosen at Princeton University, who was not involved in the work.

The world’s largest aircraft breaks cover in Silicon Valley, it's backed by Sergey Brin and is an airship (Techcrunch)

At 124.5 meters long, Pathfinder 1 dwarfs the current Goodyear airships and even the massive Stratolaunch plane designed to launch orbital rockets. It’s the largest aircraft to take to the skies since the gargantuan Hindenburg airship of the 1930s.
LTA’s airship uses stable helium rather than flammable hydrogen as a lifting gas
Can go up to 120km/h, can carry about four tons of cargo in addition to its crew
The CEO: “I can’t see airships replacing aircraft,” he said. “But I do see a niche for airships to be part of the transportation architecture that reduces the carbon footprint of air travel.” Another important niche could be responding to natural disasters like earthquakes, volcanic eruptions and hurricanes.

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.

Quelques mots sur le cuistot

J'ai écrit plus de 50 articles ces dernières années, à retrouver ici, dont une bonne partie publiés dans des médias comme le Journal du Net (mes chroniques ici), le Huffington Post, L'Express, Les Échos.
Je suis CEO et co-fondateur de l'agence digitale KRDS, nous avons des bureaux dans 6 pays entre la France et l'Asie, ainsi que de Yelda, “A voice assistant to answer every call” (notre itw sur BFM Business “À la mairie de Plaisir, c'est l'intelligence artificielle qui répond au téléphone”), une des Future 40 startups de Station F, notre étude de cas vidéo bluffante d’1 minute avec Best Western.
Retrouvez ici mon podcast Parlons Futur (ou taper "Parlons Futur" dans votre appli de podcast favorite), vous y trouverez entre autres des interviews et des résumés de livres (j’ai notamment pu mener un entretien avec Jacques Attali, un autre avec Pierre Bellanger, fondateur et CEO du groupe Skyrock sur Apple Podcast et Spotify).
Je suis basé à Singapour (mon Linkedin, mon Twitter), également membre du think tank NXU.

Merci, et à bientôt !

Thomas

Parlons Futur