AI Summit in Paris highlights global disunity on AI - Sync #506
Plus: Musk offers to buy OpenAI; OpenAI reveals its future plans; Anthropic's new model is imminent; Apple is exploring humanoid robots; biotech's DeepSeek moment; biohybrid robots; and more!
Hello and welcome to Sync #506!
We've had yet another week where news from the tech world just didn’t stop coming…
For this week's main story, we’ll take a look at the AI Summit in Paris, which highlighted global disunity as AI increasingly becomes a geopolitical tool.
Elsewhere in AI, Elon Musk surprised everyone by offering to buy OpenAI for $97 billion. Unsurprisingly, OpenAI rejected the offer and revealed that the new GPT-4.5 and GPT-5 models, as well as a new hybrid AI model, are set to be released in the coming weeks or months. Similarly, Anthropic is expected to launch its own hybrid AI model very soon.
Over in robotics, reports have emerged that Apple is exploring humanoid robots. Meanwhile, Apptronik has raised $350 million to further develop its humanoid robots, and we’ll meet researchers using paleo-inspired robotics to understand how prehistoric animals moved.
In biotech, The Wall Street Journal warns of a potential “DeepSeek moment” for US pharma and biotech companies. In Argentina, researchers have created the world's first genetically edited horses, and we also have two new advancements in biorobotics and using living cells to build robots.
Enjoy!
AI Summit in Paris highlights global disunity on AI
This week, the Artificial Intelligence Action Summit in Paris brought together global leaders to discuss the rapidly evolving landscape of AI, yet it underscored the deep divisions and competing visions shaping its future.
The summit revealed a central tension between the US's ambition to maintain dominance in AI and Europe's push for strategic autonomy. JD Vance, the US Vice President, asserted an "America First" approach to AI, signalling that the US would prioritise its own AI development and resist international regulations that could hinder its progress. Vance also cautioned against collaboration with "authoritarian" regimes—a clear reference to China—further highlighting geopolitical fault lines.
Meanwhile, Chinese Vice-Premier Zhang Guoqing indicated China’s willingness to collaborate on AI security and share its advancements to build "a community with a shared future for mankind."
For Europe, the summit was an opportunity to develop a unified strategy to establish itself as a key player in the global AI scene. The EU unveiled the InvestAI initiative, aiming to mobilise €200 billion in AI investments across Europe. Additionally, French President Emmanuel Macron announced a substantial €109 billion investment in France's AI ecosystem.
Discussions also centred on the potential economic and ethical implications of AI. Dario Amodei, CEO of Anthropic, cautioned that advanced AI could dramatically alter the global labour market. Others raised concerns about the environmental sustainability of AI and warned of AI's potential to exacerbate economic inequality.
Experts like Yoshua Bengio and Demis Hassabis have stressed the importance of addressing AI safety concerns, including the potential for deceptive AI systems. However, safety has not been at the forefront of discussions. Instead, it seems that the global conversation about AI is now shifting from “safety” towards “security.” An example of this is the recent name change of the UK’s AI Safety Institute to the AI Security Institute. With this change, the body will shift from primarily exploring areas such as existential risk and bias in large language models to focusing on cybersecurity, specifically “strengthening protections against the risks AI poses to national security and crime.” Those concerned about the existential risks posed by AI may find this shift towards security disappointing. However, those who view such concerns as “sci-fi fantasy” and advocate for a more pragmatic approach to AI safety will welcome this change.
Similarly to the previous two AI summits in the UK and South Korea, the culmination of two days of discussions was the signing of the summit declaration, and the same was true for the summit in Paris. The summit emphasised bridging digital divides, promoting ethical and trustworthy AI, and fostering innovation while ensuring AI benefits both people and the planet. Key priorities included AI accessibility, responsible governance, sustainable development, and global cooperation. Actions launched include a Public Interest AI Platform & Incubator, discussions on AI and energy sustainability, and observatories to monitor AI's impact on labour markets. The summit reaffirmed the need for multistakeholder AI governance, international safety standards, and alignment with global AI initiatives such as the Bletchley Park and Seoul Summits.
Sixty-two countries and organisations signed the declaration; however, the US and the UK did not.
The US disagreed with the terminology around multilateralism and international collaboration. The US delegation also raised concerns that Current AI, the foundation launched by France during the summit, might be used to direct funds towards French-speaking countries. Meanwhile, the UK did not sign the declaration because it felt the document did not go far enough in addressing the global governance of AI and its impact on national security. According to Ars Technica, one person close to the British government suggested the wording was "too restrictive."
One thing that became clear from the summit in Paris is the shift in how AI is perceived. At previous summits, discussions focused on making AI development and applications open, safe, and trustworthy. The language was cooperative. However, the AI Summit in Paris has shown that we are in a world where AI is also a geopolitical tool and will be used to advance national interests.
If you enjoy this post, please click the ❤️ button or share it.
Do you like my work? Consider becoming a paying subscriber to support it
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🦾 More than a human
A Year of Telepathy
Neuralink shared an update on its ongoing PRIME Study, revealing that three participants—Noland, Alex, and Brad—have received the Link, the company’s brain-computer interface (BCI). According to Neuralink, the participants have collectively used the Link for over 4,900 hours across more than 670 days, with independent use averaging 6.5 hours per day. The update also highlights the device’s impact on their lives, enabling them to control computers using only their thoughts, rediscover their creativity, and reconnect with friends and family. Additionally, Neuralink hinted at possible clinical trials in the UK, alongside the ongoing trials in the US and Canada.
Anti-ageing jabs – they can rejuvenate mice, but will they work on humans?
Senescent cells—also known as "zombie cells"—play a crucial role in ageing and chronic diseases, particularly in childhood cancer survivors who experience accelerated ageing due to harsh treatments. Scientists are investigating senolytic drugs such as dasatinib, quercetin, and fisetin, which can eliminate these cells and potentially slow ageing and improve health. Clinical trials, including one at St Jude Children’s Research Hospital, are testing whether these drugs can extend lifespan and restore function in affected individuals. While promising, researchers caution that not all senescent cells are harmful, and further studies are needed to ensure safety and long-term benefits before senolytics can be widely used to combat ageing.
Brain implant that could boost mood by using ultrasound to go under NHS trial
The NHS is launching a £6.5 million (about $8.175 million) trial in the UK using a brain-computer interface (BCI) that modifies brain activity with ultrasound to improve mood. The trial, involving 30 patients, is funded by the UK’s Advanced Research and Invention Agency (Aria). The device, Forest 1, is developed by the US-based non-profit Forest Neurotech and is designed to be implanted beneath the skull but outside the brain. The study will run for 3.5 years, starting in March, with the first eight months focused on regulatory approval. If successful, this trial could revolutionise mental health treatment by offering a less invasive alternative to current brain stimulation techniques.
🧠 Artificial Intelligence
Elon Musk-led group makes $97 billion bid for control of OpenAI
Elon Musk made an unexpected move in his ongoing battle with OpenAI and Sam Altman by proposing to buy OpenAI for $97.4 billion. As TechCrunch points out, the Musk-led group behind the offer consists of investors with past or present connections to Musk’s various companies, such as Tesla and SpaceX. Altman, unsurprisingly, refused the offer and, in response, proposed to buy Twitter for $9.74 billion. OpenAI then reiterated its decision to reject the offer in a letter to staff. Later in the week, a full letter of intent was made public, revealing that Musk’s offer comes with a specific expiration date: 10 May 2025. It would be an all-cash transaction, with the buyers demanding full access to OpenAI’s financial records and personnel for due diligence before completing the acquisition. However, in a new twist, Musk’s legal team stated that he would drop his bid to acquire OpenAI if the board commits to keeping the organisation as a nonprofit. Some analysts argue that the offer was made to complicate the process of turning OpenAI into a full for-profit company by driving up the price Altman would have to pay to take the company private.
OpenAI lays out plans for GPT-5
Sam Altman has revealed in a tweet OpenAI’s plans for the near future. The company intends to release GPT-4.5 (internally known as Orion) as its last non-chain-of-thought model. The long-term plan is to develop a hybrid model—similar to what Anthropic is working on (more on that later)—that combines o-series models and GPT-series models, and knows when to use faster GPT models and when to take time to reason through a problem. For both ChatGPT and OpenAI’s API, the company plans to “release GPT-5 as a system that integrates a lot of our technology, including o3,” Altman says, adding that “we will no longer ship o3 as a standalone model.” While no concrete release dates have been given, Altman hinted at timelines measured in weeks or months.
Anthropic’s next major AI model could arrive within weeks
According to a report from The Information, Anthropic is preparing to launch a major new model soon. The new model reportedly outperforms OpenAI’s o3-mini-high model on some programming tasks and excels in analysing large codebases and other business-related tasks. Additionally, Anthropic’s new model is rumoured to exhibit hybrid capabilities, switching between “deep reasoning” and fast response modes. Dario Amodei suggested that Anthropic’s goal is to improve reasoning models and questioned the industry distinction between "normal" and "reasoning" models.
OpenAI co-founder Sutskever's SSI in talks to be valued at $20 billion
Safe Superintelligence (SSI), an AI startup co-founded by former OpenAI chief scientist Ilya Sutskever, is seeking funding at a valuation of at least $20 billion—four times its previous $5 billion valuation from its $1 billion funding round in September 2024. According to Reuters, fundraising is still in the early stages, and the exact amount SSI is seeking is unclear.
Lyft to bring Claude to more than 40 million riders and over 1 million drivers
Lyft is partnering with Anthropic to develop AI-powered products that enhance the rideshare experience for both riders and drivers. This initiative involves early research, model testing, and engineering improvements. According to the statement, the partnership is already impacting customer experience by reducing customer service resolution time by 87% and handling thousands of daily inquiries.
Macron unveils $112B AI investment package, France’s answer to US’ Stargate
French President Emmanuel Macron announced ahead of the Artificial Intelligence Action Summit in Paris a €109 billion ($112 billion) investment package to accelerate AI innovation in France and Europe. The majority of the funds will be used to build AI-focused data centres, aligning with global trends such as OpenAI’s Stargate Project in the US.
Mistral CEO: Europe must ‘own and operate’ its AI infrastructure
Speaking at a panel at the AI Action Summit in Paris, Mistral CEO Arthur Mensch urged Europe to invest more in AI infrastructure to avoid falling behind the US and China. Mensch emphasised that Europe should develop and own AI infrastructure to prevent profits from flowing back to US-based hyperscalers. Prominent tech figures, including DeepMind’s Demis Hassabis, LinkedIn’s Reid Hoffman, and Anthropic’s Dario Amodei, echoed Mensch’s concerns. Former Google CEO Eric Schmidt, who also attended the summit, stated that Europe's ambition in AI is equal to or greater than that of the US, but structural issues hinder progress. Founders and investors at the conference urged the EU to implement regulations that encourage innovation while maintaining competitiveness.
Veo 2 is coming to YouTube Shorts
Veo 2, Google DeepMind's video generation model, is coming to YouTube Shorts, allowing creators to add high-quality videos to their shorts for “an easy and fun way to express themselves,” according to YouTube's announcement. The new feature will initially be available in the US, Canada, Australia, and New Zealand, with plans to expand to more regions in the future.
Three Observations
In this blog post, Sam Altman outlines three observations about AI: (1) the intelligence of an AI model scales predictably with increasing computational resources, (2) AI costs are decreasing rapidly, making AI more accessible—Altman cites the example of GPT-4o’s token cost falling 150x from 2023 to 2024., and (3) the economic value of AI increases exponentially with intelligence, driving continued investment. He then envisions a near future in which AI could be as transformative as the transistor, embedding itself into every aspect of the economy.
▶️ DeepSeek’s Lessons for Chinese AI (16:41)
In this video, Jon from Asianometry breaks down DeepSeek R1, exploring its innovative and efficiency-focused architecture, as well as the company's unique (for a Chinese company) approach to hiring and nurturing talent. Jon raises key questions about whether DeepSeek can maintain its research-driven culture and retain brilliant employees as it goes mainstream. He also examines how the company will tackle challenges related to scaling up its product and securing sufficient compute power to compete with top Western and Chinese rivals.
Google starts testing new Search ‘AI Mode’ internally – Here’s an early look at it
Google Search is testing a new “AI Mode” powered by Gemini 2.0, enabling open-ended and exploratory queries. This AI-powered Search is designed for complex, research-based questions that the current Search struggles with. According to 9to5Google, the new Search is currently being tested with Google employees in the US and may be publicly released this year.
Meta in talks to acquire South Korean AI chip startup
Meta is reportedly in talks to acquire FuriosaAI, a South Korean AI chip startup. According to Forbes, the discussions are at an advanced stage, and the deal could be signed by the end of February. FuriosaAI develops AI inference accelerators for data centres, with its latest chip, RNGD (pronounced “renegade”), claiming to offer better cost and energy efficiency than Nvidia’s high-end processors.
Thomson Reuters wins AI copyright 'fair use' ruling against one-time competitor
A US federal judge in Delaware ruled that Ross Intelligence, a defunct legal research firm, violated US copyright law by copying Thomson Reuters' content to build an AI-based legal platform. This marks the first US ruling on the fair use doctrine in AI-related copyright cases, setting a precedent for similar litigation.
▶️ Deep Dive into LLMs like ChatGPT (3:31:23)
Andrej Karpathy, a well-respected AI researcher and now an AI educator, has dropped a three-hour-long video that is a perfect introduction to large language models. It covers all essential topics, from pre-training and tokenisation to building a model, training, post-training, and the latest developments in the field. I highly recommend this video to anyone interested in understanding how these models work and what to expect in the near future.
AI ‘godfather’ predicts another revolution in the tech in next five years
Yann LeCun, Meta's chief AI scientist, has stated that current AI systems are too limited to enable domestic robots and fully automated cars, requiring new breakthroughs to understand and interact with the physical world. He predicts that such a breakthrough should occur by the end of the decade.
The Anthropic Economic Index
Anthropic is launching the Economic Index—an initiative aimed at understanding AI's effects on labour markets and the economy over time. The company has also released the first report from the Index, which found that AI is most used in software development and technical writing, though the authors note that coding might be overrepresented in AI usage data. The report also revealed that AI is used more for augmentation (57%) than automation (43%) and that mid-to-high wage roles (e.g., programmers, data scientists) show higher adoption of AI than both low- and high-paid roles. Anthropic promises to regularly release results and associated datasets as part of the Anthropic Economic Index. The full report can be found here.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
🤖 Robotics
Apple is reportedly exploring humanoid robots
According to Ming-Chi Kuo, Apple is exploring both humanoid and non-humanoid robotic form factors. According to his scoop, Apple’s humanoid robot project is in the "early proof-of-concept" stage, and it is still unclear whether it will become an actual product or share a fate similar to the now-abandoned Apple Car project. If the project goes ahead, mass production is expected to start in 2028 at the earliest, writes Kuo. Given Apple's development timeline, at least three more years of leaks and speculation are expected before any real product emerges.
Apptronik, which makes humanoid robots, raises $350M as category heats up
Apptronik, a University of Texas spinout building humanoid robots, has raised $350 million in a Series A round led by B Capital and Capital Factory, with participation from Google. Previously, the startup had raised a modest $28 million and now aims for commercialisation and scaling in 2026 and beyond. Apptronik’s humanoid robotics work dates back to 2013, before its official founding in 2016. It originated from the DARPA Robotics Challenge, working on the Valkyrie humanoid robot. Currently, Apptronik has ongoing pilots with Mercedes and GXO Logistics, but none have moved beyond the pilot stage yet. The company has also partnered with Google DeepMind to develop robot behavioural models.
Robots are bringing new life to extinct species
Paleo-inspired robotics is one of the coolest applications of robotics I’ve seen. To better understand how prehistoric animals moved, researchers build robots resembling these animals and test their theories. In other words, these researchers are literally building robot dinosaurs.
Machine learning transforms mini biohybrid ray design, doubling swimming efficiency
You may have heard about a biohybrid stingray robot built with rat heart cells. It was an amazing achievement in biorobotics, blurring the line between robots and living organisms. Now, a group of researchers has applied machine-learning-directed optimisation (ML-DO) to improve biohybrid robot design. The result is a new robot design that uses large aspect ratio fins with finely tapered tips, leading to higher efficiency across different swimming scales. When researchers built these improved biorobots, they found that these robots demonstrated double the swimming efficiency compared to previous biomimetic designs.
🧬 Biotechnology
The Drug Industry Is Having Its Own DeepSeek Moment
The release of DeepSeek has severely shaken US dominance in AI. As The Wall Street Journal reports, the same is happening in the biotech and pharma industries. In 2020, less than 5% of large pharmaceutical deals ($50M+ upfront) involved Chinese companies. By 2024, that figure had risen to nearly 30%. Many drugs entering the US market in the future will originate from Chinese labs. Chinese biotech companies operate with lower costs and greater speed than their US counterparts, benefiting from efficient regulatory processes and cheaper clinical trials. While Chinese innovation is currently incremental—focusing on improving existing drugs—it is becoming more disruptive. If this trend continues, China’s biotech sector will soon challenge US dominance and reshape the global landscape of biotech innovation.
Argentina's gene-edited horses
Scientists from Argentina have created the world's first genetically edited horses. Using CRISPR, they have taken DNA from a polo champion mare, Polo Pureza, and modified it to enhance "explosive speed" in her offspring. Five foals were born in October and November with mostly the same genes as their award-winning predecessor. Kheiron Biotech, the firm behind the project, claims the changes could have occurred naturally via selective breeding—CRISPR only helped accelerate nature’s processes. However, at the moment, these genetically engineered horses won’t be allowed in international competitions, as the International Federation for Equestrian Sports bans gene-edited horses (though it allows clones).
Biohybrid hand uses sushi-like rolls of lab-grown human muscle to move objects
A team from the University of Tokyo and Waseda University in Japan has created a biohybrid hand that can perform gestures, including a scissor motion, and manipulate objects. The hand has been built using lab-grown muscle tissue strands called MuMuTAs. These are thin muscle fibres bundled together like sushi rolls to provide enough strength for contraction. The hand successfully performed a scissor gesture and manipulated a pipette tip. However, muscle fatigue occurs after 10 minutes of use (it recovers in one hour), and the hand must be suspended in liquid to avoid friction and allow smooth movement. There is still room for improvement, but MuMuTAs mark a major step towards realistic, functional biohybrid limbs.
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!
My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"