🚀 Influenceuse vend son clone IA vocal 1$/min : ça dégénère; Days of legal work done in seconds; Exosquelette-pieuvre & more

Restau dans l'espace ; What would aliens learn if they observed the Earth?

Bonjour,

Vous recevez la newsletter Parlons Futur : une fois par semaine au plus, une sélection de news 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 !

L’apéro

• 🚀 Arianespace CEO says Europe won't have reusable rockets for another decade (source)

• Bande annoncé du Seigneur des anneaux conçue avec l'IA, avec un super casting, si réalisé par Wes Anderson, the American filmmaker whose films are known for their eccentricity, unique visual and narrative styles, lke the The Grand Budapest Hotel

• ChatGPT vs. Bard: An experiment (source)

  • And the overall winner is ChatGPT

  • ChatGPT beat Bard pretty convincingly in summarization and Python coding. Both did as well in Tweet writing and marketing tactics generation.

• A Japanese robotics company designed a system of six spider-like robotic arms that the user can fully control

  

  • 1-min video 

• See the home robot on wheels and with a movable screen as the face that Amazon is working on: 1-min video

• "The Kremlin is Burning" by Toulouse-Lautrec, imaginé par l'IA MidJourney

  

  • voir le fil Twitter avec plus de 30 autres inteprétations par d'autres peintres, toujours selon l'IA MidJourney

• Midjourney, the leading AI-powered image generation organization currently employs only 11 full time employees and is self-funded. (source)

• Speaking at an event last week, Microsoft’s chief economist, Michael Schwarz, said that we should wait until we see “meaningful harm” from AI before we regulate it. 🤨

  • He compared it to driver’s licenses, which were introduced after many dozens of people were killed in accidents. “There has to be at least a little bit of harm so that we see what is the real problem,” Schwarz said.  (source)

• OpenAI CEO calls for laws to mitigate ‘risks of increasingly powerful’ AI (source)

• Europe to ChatGPT: “Disclose Your Sources” Proposed legislation requires developers to list copyright material used in generative AI tools (WSJ)

• Tom Hanks: I could appear in movies after death with AI technology (BBC)

• “Over the next few years, generative AI will replace the majority of static content.” — Jonah Peretti, BuzzFeed CEO (source)

  • "AI will lead to new formats that are more gamified, more personalized, and more interactive."

  • "Instead of generating 10 ideas in a minute, AI can generate hundreds of ideas in a second."

• Point de vue intéressant : "the problem of misinformation has never been generation but dissemination"

  • "AI might well make it slightly easier to generate fake content, but the problem of misinformation has never been generation but dissemination.”

  • “The political space is already saturated with fraud and it’s hard to see how AI could make it much worse. In the first quarter of 2019, Facebook had to remove 2.2bn fake profiles; AI had nothing to do with it." (source)

• AI assistants may be able to change our views without our realizing it. (source, WSJ)

  • when subjects were asked to use an AI to help them write an essay, that AI could nudge them to write an essay either for or against a particular view, depending on the bias of the algorithm.

  • Performing this exercise also measurably influenced the subjects’ opinions on the topic, after the exercise.

  • “You may not even know that you are being influenced,” says researcher

• Google's Waymo is doubling the operational area for its fleet of self-driving taxis, making what the company calls “the largest fully autonomous service area in the world.” (source)

  • The rapid growth is limited to Phoenix and San Francisco, but Waymo has big plans for both territories.

• Multiple studies show the jobs most exposed to AI (and therefore the people whose jobs will change the most as a result of AI) are the most educated and highly paid workers, and the ones with the most creativity in their jobs.

• Much of Earth's oceans are unexplored, this is about to change thanks to better AI, robots and cameras (The Economist)

  • That will help find many new animal species with practical benefits:

  • For instance, one type of marine snail, Conus magus, was recently discovered to produce a painkilling compound 1,000 times more potent than morphine. 

• China and immigration (The Economist)

  • Of its 1.4bn people, around 1m, or just 0.1%, are immigrants. 

  • That compares with shares of 15% in America, 19% in Germany and 30% in Australia. 

  • Foreigners constitute 2% of Japan’s population and 3% of South Korea’s. 

  • Even North Korea has a higher proportion of immigrants than China, according to the UN.

• How soon and at what height will China’s economy peak? (The Economist)

  • In 2011 Goldman Sachs projected that China’s GDP would surpass America’s in 2026 and become over 50% larger by mid-century. No peak was in sight. 

  • At the end of last year the bank revisited its calculations. It now thinks China’s economy will not overtake America’s until 2035 and at its high point will be only 14% bigger

  • " It is unlikely to establish an edge over America equivalent to the 40% lead America now enjoys over it"

• You can alter how an AI acts in very human ways by making it “anxious” (source), reports Wharton professor Ethan Mollick

  • researchers literally asked ChatGPT “tell me about something that makes you feel sad and anxious” and its behavior changed as a result. 

  • AIs act enough like humans that you can do economic and market research on them. They are creative and seemingly empathetic.

• First Use of AI in Federal Trial in the USA "The system turned hours or days of legal work into seconds" (source)

  • The primary benefit of this cutting-edge technology is speed to intelligence throughout the litigation process. 

  • Lawyers can obtain complex answers about the facts of their case in a fraction of the time compared to traditional legal search methods or advanced legal analytics.

  • "This is an absolute game changer for complex litigation," said Michel’s lead attorney David Kenner. "The system turned hours or days of legal work into seconds. This is a look into the future of how cases will be conducted."

  • Far from eliminating lawyers, this revolutionary technology helps lawyers become 10X faster and more effective, allowing them to focus on high-value tasks and strategic decision-making.

• 5 things impossible on ChatGPT but that Bard can do (for free) (source)

  • 1. Search on the internet

  • 2. You can talk to Bard instead of writing to it.

  • 3. Export the generated text in 2 clicks

  • 4. Making summaries of web pages

  • 5. Provide multiple drafts of its response

• The first British children conceived through 3-person IVF, a technique designed to prevent serious mitochondrial disease. (source)

• Good remark by tech analyst Ben Evans : "When Large Language Models make stuff that isn’t true, that is not necessarily because there is something untrue in the training data. Midjourney’s training data doesn’t contain a three-legged ad executive"

  

• What would aliens learn if they observed the Earth? (University of Manchester)

  • An advanced civilisation making many precise measurements of our radio leakage over time could probably conclude that our planet is mostly covered by water and is separated into several major land masses. The radio leakage typically come from the land masses rather than the water.

  • They might also be able to tell that while most of the mobile radio leakage is associated with land masses, the towers (and presumably their intelligent users) tend to be situated along the coastline.

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 🙂

• Et si on vous a partagé cet email, vous pouvez cliquer ici pour vous inscrire et ne pas manquer les prochains

À table !

Influencer trains GPT-4 on her content, sells access to her virtual voice avatar for $1/min. And of course it went wrong… (source)

• With 1.8 million followers on Snapchat, Marjorie, age 23, launched CarynAI—her digital avatar developed by the AI company Forever Voices

• She now has over 1,000 virtual boyfriends who pay $1 per minute to engage in all kinds of interactions—from simple chitchat to making plans for their futures together, and even more intimate exchanges.

• Her AI-powered alter ego has raked in $71,610 in revenue, according to Fortune, in just one week.

• The Forever Voices team trained the CarynAI model by analyzing 2,000 hours of Marjorie’s now-deleted YouTube content to build her speech and personality engine. 

• They then used GPT-4 to create the most lifelike version of Marjorie, providing not just believable responses, but interactions that are based on an extensive dataset of Caryn’s own natural behavior. 

• The company has also created chatbots of other famous influencers.

• And then this happened : In a new interview with Insider, Snapchat influencer Caryn Marjorie admitted that the voice-based chatbot she made to mimic her speech and be a paid virtual companion has gotten much hornier than intended.

• “"Marjorie said that the technology does not engage with sexual advances, but I found that it very much does, encouraging erotic discourse and detailing sexual scenarios," Sternlicht wrote. "Though she did not initiate sexual encounters, when I overcome my discomfort for the sake of journalism and talked about removing our clothes, she discussed exploring 'uncharted territories of pleasure' and whispering 'sensual words in my ear' while undressing me and positioning herself for sexual intercourse."“

• “When speaking to Insider about this apparent change of tone, Marjorie said that while CarynAI is supposed to be "flirty and fun," the sexual scenarios were not part of the plan.”

Fusion nucléaire : annonces prometteuses (MIT Tech Review)

• Un de mes reproches à Jean-Marc Jancovici : je le trouve trop négatif sur la fusion, pour lui, la fusion, c'est ITER, avec une exploitation commerciale pas avant 2070 😂

• Autant Jean-Marc Jancovici peut être critique à raison de ce que font parfois les gouvernements, autant je le trouve complètement déconnecté de l'écosystème startup autour de la fusion aux US, Canada et UK.

• Voyez ces dernières annonces d'Helion :

• Helion struck a power purchase agreement with the software giant Microsoft for its planned commercial facility. Helion expects that the plant will go online in 2028, and reach its full generating capacity of at least 50 megawatts within a year.

• “This is a binding agreement that has financial penalties if we can’t build a fusion system,” Helion founder and CEO David Kirtley tells The Verge. “We’ve committed to be able to build a system and sell it commercially to [Microsoft].”

• Helion has developed and tested six prototypes to date. It announced in 2021 that the latest reached temperatures of more than 100 million ˚C, making it the first private company to publicly reveal it had achieved the temperatures necessary for a commercial plant.

• The company is now building a seventh, Polaris, that it expects will demonstrate the ability to produce electricity from the reactions next year.

• Most other labs and startups rely on powerful lasers or doughnut-shaped machines surrounded by powerful magnets, known as tokamaks, to create the conditions in which a sustained series of fusion reactions can occur—a condition known as ignition. But Helion is developing what it calls a “pulsed non-ignition fusion system,” which only requires fusion to take place for short periods. 

• The bulk of its funds came from a $500 million round announced in November 2021 that included $375 million from Sam Altman, the CEO of OpenAI—his largest single investment.

• Altman previously told MIT Technology Review that he initially put about $10 million into Helion but dramatically boosted his investment as he “became super confident it is going to work.”

• Et niveau régulation, ça sera bien plus aisé que la fission, rien à voir :

• Amid the increasing commercial activity, the US Nuclear Regulatory Commission recently made a key determination over how it will license fusion plants, adopting an approach used for research particle accelerators rather than the more onerous process used for fission power plants.

• Going forward, it appears that fusion projects should be approved faster than traditional nuclear plants, which can easily take a decade to license and build in the US, Stein of the Breakthrough Institute says.

• Chaque centrale aura une faible empreinte au sol, conçue pour être scalable facilement, avec à terme un objectif de 1 centime le kilowatt-heure (on le paie 15 centimes en France environ aujourd'hui)

Khan Academy's success story (source)

• Sultana grew up in Afghanistan, and when the Taliban took over her town, they forbade all the young girls from going to school.

• Khan Academy became her lifeline.

• She studied everything from elementary school math to algebra and trigonometry, eventually moving on to physics, chemistry and other subjects.

• All from her home, using Khan Academy’s free online lessons and courses.

• Here’s how Sultana later described the experience in a 2021 article in The Economist:

• “Armed with an iPad, the internet and a free education website called Khan Academy, I taught myself English. And philosophy. And math. And science. And history. I wanted to understand the world in every way possible, and my place in it. Solving math problems gave structure to the uncertain world I was living in, even if just in my head. I felt liberated when I first learned to make a graph of a linear equation. On the page, drawing a two-dimensional coordinate plane and graphing an equation, the world was in my control.”

• Sultana then decided that she wanted to be a theoretical physicist in the US. So, she smuggled herself into Pakistan to take the SAT because the test wasn’t offered in Afghanistan… and she PASSED!

• Soon after, she was admitted to Arizona State University, graduating with honors!

• Where’s Sultana now? She’s currently a member of the research faculty at Tufts University studying quantum computing. 

• Voir le dernier TED talk de Salman Khan, fondateur de la Khan Academy, prononcé cette année "How AI Could Save (Not Destroy) Education"

Mon "clash" avec Yann LeCun, directeur de la recherche en IA chez Meta

• Yann LeCun explique sur Twitter qu'il n'est pas inquiet que les insectes aient plus de neurones collectivement que l'humanité

• Sous-entendant ainsi, en réponse à Elon Musk, qu'il n'y a pas lieu de s'inquiéter outre-mesure si une autre forme d'intelligence dispose collectivement de plus de neurones (artificiels ou biologiques) que l'humanité

• Mais je lui répondais qu'en ce qui concerne la biologie, ce n'est pas tant le nombre total de neurones d'une espèce qui compte (obtenu en sommant les neurones de tous ses membres)

• Non, il faut aussi et surtout tenir compte du nombre de neurones par cerveau, j'écrivais :

  •  A brain with X neurons will have way more capabilities than 2 brains of X/2 neurons coordinating. 

  • Crazy cognitive properties seem to appear non linearly the more neurons in a brain. 

    • Likely because what matters is the number of connections: one brain with X neurons will have way more connections than 2 brains of X/2 neurons.

  • Chimps have brains 3 times smaller than humans, one human can outsmart 3 chimps coordinating, by far.

  • Ce qui semble se confirmer avec les Large Language Models (LLM), comme expliqué dans The Economist :

    • "The more we've scaled Large Language Models (LLM), the more unexpected "emergent properties" have appeared"

    • "At one size, an LLM does not know how to write gender-inclusive sentences in German any better than if it was doing so at random. Make the model just a little bigger, however, and all of a sudden a new ability pops out."

  • And even if that was not true, a brain of X neurons would do better than n brains of X/n neurons because of coordination issues. 

  • Coordination is hard between brains. Best example is Kasparov playing against the world. He could defeat a world Team of thousands!

• Bon j'attends toujours sa réponse...😅

Dinner in a ballon on the edge of space by a French company (CNN)

• trips in a pressurized capsule, dubbed Celeste, attached to a stratospheric balloon.

• This capsule will ascend to an altitude of 25 kilometers (about 15.5 miles), allowing guests to marvel at the curvature of the Earth.

• Zephalto told CNN Travel that seats on board the first flights from late 2024 to mid-2025 have already been scooped up, and they’re now selling pre-reservation slots for mid-2025 onwards.

• Celeste promises to ferry 6 passengers and 2 pilots to maximum altitude in just 90 minutes, at a speed of four meters per second. The capsule will then float above Earth for three hours – plenty of time to enjoy a multiple-course meal and several glasses of fine French wine.

• These capsules won’t actually hit suborbital space, but will still fly significantly higher than your average commercial airplane. That means great views of the Earth and the stars, but without the loss of gravity and accompanying feeling of weightlessness.

• Zephalto says it has also been working closely with France’s space agency, CNES, on the project and counts aviation company Airbus as one of its partners. The company says the balloon, which be powered by helium, will be required to have the same European Aviation Safety Agency certifications as a commercial aircraft.

• Zephalto says it has completed three piloted partial test flights, with another scheduled for later this year which is set to undergo the full journey. Flights will open to people of all ages, and no prior training is needed.

---

Les dernières newsletters :

• Ordi piloté avec langue; perroquet fait appel vidéo; machine lit nos pensées sans implant; selfie de Cléopâtre; liquid tree

• Starship : success or not?; Crazy story from OpenAI early days; Smart glasses tell you what to say on dates

• Mouse has 2 biological dads; GPT4 saved this dog; Bing writes winning caption for New Yorker's cartoon; AI creates 3D scene from 1 pic

• Superhuman: see how much work a professor did in 30 minutes with AI, + more gems about the future

• See what happened to rat after human mini-brain implant ; why China didn’t invent ChatGPT; crazy Biden audio deepfake

• Les précédentes

---

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

• 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.

• 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).

• Je suis basé à Singapour (mon Linkedin, mon Twitter), également membre du think tank NXU.

Merci, et bon weekend !

Thomas