Humanoid robots for factories and for homes - Sync #483
Plus: new rumours about OpenAI's new models; DeepMind AlphaProteo; DIY, pirated medicine; OpenAI and Anthropic to share models with US government; and more!
Hello and welcome to Sync #483!
This week, we’ll take a closer look at the recent developments in humanoid robotics. We’ll explore the plans to make humanoid robots a new engine of economic growth in China, Figure’s successful completion of trials with BMW, and 1X’s announcement of a humanoid robot for the home.
In other news, Ilya Sutskever’s new AI venture has raised $1 billion to create safe superintelligence. Meanwhile, OpenAI and Anthropic will be sharing their new models with the US government for safety evaluations and xAI’s supercomputer for AI is now online. We also have fresh rumours about OpenAI’s upcoming models.
Over in robotics, researchers have created a biohybrid robot controlled by mushrooms, and a window-cleaning robot has been deployed in New York.
We’ll finish this week’s issue of Sync with AlphaProteo—DeepMind’s new AI system that designs novel proteins, and with a story about a group that makes DIY pirated medicine.
Enjoy!
Humanoid robots for factories and for homes
China enters the humanoid robotics race
In 2023, China’s Ministry of Industry and Information Technology (MIIT) outlined its ambitious goals for developing humanoid robots. According to the plan, China aims to be ready to mass-produce humanoids by 2025. The plan calls for establishing a humanoid innovation system, making breakthroughs in several key technologies, and ensuring the safe and effective supply of core components. The document states that by 2027, humanoid robots should become an important new engine of economic growth in China.
It’s been two years since MIIT outlined its plans and we can see its results. At the recent World Robot Conference in Beijing, 27 Chinese robotics companies presented their humanoid robots.
China’s plan with humanoid robots is to achieve the same results as it did with EVs and other tech products. This plan includes government support, ruthless price competition from a wide field of new entrants, and a deep supply chain. We can already see this plan unfolding.
The government is investing billions into robotics. The city of Beijing launched a $1.4 billion state-backed fund for robotics in January, while Shanghai announced plans in July to set up a $1.4 billion humanoid industry fund.
Those 27 humanoid robotics companies represent the second pillar of the Chinese robotics plan—the wide field of new entrants that will ruthlessly compete with each other.
Initially, these Chinese companies will target the domestic market, learning what works and what does not. They will quickly iterate and innovate their products on a scale larger than in the West, with access to one of the world’s best supply chains, helping them keep prices as low as possible. Eventually, the remaining companies will have highly competitive and likely cheaper products than their Western counterparts. Once that happens, customers from outside China will show up.
This playbook has been applied to EVs and it resulted in BYD and Nio, Chinese EV companies, outselling Western competitors. Now, the same playbook is being applied to humanoid robots.
One of the leading Chinese robotics companies is Unitree, which offers two humanoid robots—G1 and H1. Recetnly, G1 has been announced to be ready for mass production. Also, Unitree is the only company so far that revelead the price for their humanoid robots. G1’s price starts at $16,000. The price for the larger robot, H1, is listed as $90,000 but the website asks to contact sales for “real price.”
Figure’s humanoid robot successfully completes trials with BMW
Earlier this year, Figure, one of the leading companies in this new wave of humanoid robotics companies, announced that their robot has started trials with BMW at the carmaker’s factory in Spartanburg, South Carolina.
This week, BMW announced in a press release that the trials have been successful. Figure’s robots have been tested in a real production environment, where they autonomously performed complex tasks such as inserting sheet metal into fixtures for chassis assembly. Both companies will continue collaborating to develop these robots for future use in production environments.
“The developments in the field of robotics are very promising. With an early test operation, we are now determining possible applications for humanoid robots in production. We want to accompany this technology from development to industrialization,” says Milan Nedeljković, Member of the Board of Management for Production at BMW AG.
Figure is the first humanoid robot to successfully complete trials with a carmaker, but there are others still being evaluated. Earlier this year, Mercedes began testing Apptronik’s Apollo humanoid robots for 'low-skill, repetitive' tasks. Meanwhile, Tesla is planning to introduce its humanoid robot, Optimus, into its factories next year.
1X NEO Beta—a humanoid robot for home
Meanwhile, 1X is taking a different path and, instead of bringing humanoid robots to factories and warehouses, they plan to bring these robots into homes. The Norwegian startup announced last week that its humanoid robot, NEO, is entering the pre-production phase. 1X said it is preparing for pilot deployments in selected homes by the end of this year.
Building a humanoid robot is already a challenge, but building one that operates in homes, close to people, takes that challenge to a new level. NEO was designed from the ground up to be a consumer robot, the company says. For that, the robot is designed to be lighter and softer than its competitors, which are built with industrial applications in mind.
1X did not disclose how much NEO could cost, but I wouldn’t be surprised if the price tag is in the tens of thousands of dollars. Alternatively, the robot could be provided using a Robot-as-a-Service (RaaS) model, where a customer pays a monthly fee to use the robot.
But just like with industrial humanoid robots, the expectations are high. For decades, science fiction has been (and still is) painting a picture of a future where home humanoid robots take care of chores while humans focus on other things. For industrial humanoid robots, the expectation is that they can be dropped into a warehouse or factory floor and start working after a very short period of training. We are not there yet for either home or industrial scenarios, and the first contact these robots have with reality could be crucial.
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 Logo on a Prosthesis Is Like a Tattoo You Didn’t Ask For
This article asks an intriguing question—should bionic prosthetics have their manufacturer’s logos on them? The author shares their and other bionic limb users' experiences with having various logos imprinted on what the manufacturers see as a product, but they see as parts of their bodies.
🧠 Artificial Intelligence
OpenAI and Anthropic will share their models with the US government
OpenAI and Anthropic have agreed to grant the US AI Safety Institute early access to their AI models to improve safety, assess risks, and provide feedback before and after model releases, aligning with broader efforts to create AI safety measures without stifling innovation. US AI Safety Institute director Elizabeth Kelly said in a statement that the new agreements were “just the start, but they are an important milestone as we work to help responsibly steward the future of AI.”
Ask Claude: Amazon turns to Anthropic's AI for Alexa revamp
Amazon is reportedly working on a new version of Alexa, set for release in October, which will use Anthropic’s Claude model instead of in-house AI models. The new version, dubbed "Remarkable Alexa," will be a paid service costing $5 to $10 per month, offering advanced generative AI capabilities to handle more complex tasks. While the classic, free Alexa will remain available, Amazon hopes the paid upgrade will help boost revenue and keep pace with competitors like ChatGPT.
OpenAI Considers $2,000 Monthly Subscription Prices for New LLMs
According to a report from The Information, OpenAI executives were discussing subscription prices ranging up to $2,000 per month for the upcoming Strawberry and Orion models. These models are reportedly able to solve problems and tasks that are beyond the capabilities of current AI models, as well as being capable of doing “deep research.” OpenAI’s new model, Strawberry, is rumoured to be released by the end of this year.
"ChatGPT has 200 million active users" - OpenAI Japan representative announces, also mentions next-generation model "GPT Next"
Speaking at a business event in Japan, Tadao Nagasaki, the CEO of OpenAI Japan, said that the number of active ChatGPT users had exceeded 200 million as of the end of August, the fastest software in history to reach 100–200 million active users. He also mentioned that OpenAI is working on a new model named GPT Next. According to one slide from his presentation, this new model is supposed to be released quite soon and be much more capable than OpenAI’s current best model, GPT-4. The link above points to a translated article—the original Japanese article can be found here.
xAI’s AI supercomputer is online
Elon Musk has announced that xAI’s massive supercomputer for training AI models, named Colossus, is now online. The supercomputer consists of 100,000 Nvidia H100 GPUs, and it took 122 days to complete. However, Musk says that an even bigger supercomputer is in the works and should be online in a couple of months.
OpenAI co-founder Sutskever's new safety-focused AI startup SSI raises $1 billion
Safe Superintelligence (SSI), Ilya Sutskever’s new company, has raised $1 billion in cash to help develop safe artificial intelligence systems that far surpass human capabilities. Investors include Andreessen Horowitz, Sequoia Capital, DST Global, and SV Angel. NFDG, an investment partnership run by Nat Friedman and SSI's Chief Executive Daniel Gross, also participated. Sutskever said his new venture made sense because he "identified a mountain that's a bit different from what I was working on."
Workers at Google DeepMind Push Company to Drop Military Contracts
Nearly 200 Google DeepMind employees signed a letter urging the company to cease its contracts with military organisations. The letter argues that military involvement undermines Google's AI Principles, which prohibit developing AI for harmful applications or those that violate international law. Google maintains that its AI complies with ethical guidelines, but employees remain unsatisfied with these explanations.
German LLM maker Aleph Alpha pivots to AI support
Aleph Alpha, a German AI startup developing its own large language model, appears to concede and pivot to offering support and expertise to those who want to use AI in their businesses or organisations. The startup, which raised over $640 million, was marketing itself as the “European LLM,” but, as the company’s CEO Jonas Andrulis said, that is “not sufficient as a business model.” This news is another sign that the LLM market, which was booming with investments and new startups after the release of ChatGPT in November 2022, is entering the consolidation phase.
AI-Implanted False Memories
MIT researchers published a study investigating the role of AI in inducing false memories during simulated crime witness interviews. Their results show that chatbots significantly increased false memory formation, inducing over three times more false memories than the control group. Their findings raise ethical concerns about AI use in sensitive contexts like police interviews.
▶️ How might LLMs store facts (22:42)
In this new video from their Neural Networks series, 3Blue1Brown explains how large language networks store facts. Their explanation is quite approachable, although some basic knowledge of linear algebra will be helpful (3Blue1Brown also has a linear algebra course if you want to learn it). In any case, it is a good video that will leave you with a better understanding of how large language models work. Additionally, it explains why the performance of large language models scales so well with the size of the model.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
🤖 Robotics
First-world problems drives Apple's robotics development
Apple is reportedly exploring robotics as a new avenue for expansion, as breaking into existing categories is becoming more difficult. The company has considered concepts like a robotic arm with an iPad attached, and hardware that could assist with solving "first-world problems." The effort is headed by Kevin Lynch, VP of Technology at Apple, with the company hiring more robotics experts and considering AI-driven interfaces that may go beyond Siri. While ambitious, these projects are still in the early stages, with a potential release in 2026 or 2027.
Biohybrid robots controlled by electrical impulses — in mushrooms
Researchers from Cornell University have created robots that are controlled by fungal mycelia. These biohybrid robots use electrical impulses from fungal mycelia to control their movements and can respond to environmental stimuli, such as light and chemicals. This work offers a new path for designing robots that incorporate living systems to interact with their surroundings more organically compared to traditional robots.
Skyline Robotics deploys Ozmo window cleaning robot in New York City
There are now robots cleaning the windows of skyscrapers in New York. Created by Skyline Robotics, Ozmo is a pair of robotic arms using the same platforms human cleaners use. According to the company, the robots can autonomously clean windows three times faster than humans.
Improved version of ‘Robocrop’ only picks ripe raspberries
Fieldwork Robotics has released a new version of their raspberry-picking robot, Fieldworker 1, nicknamed Robocrop (excellent name by the way). The company says Robocrop is capable of matching human speed and accuracy in harvesting and is already being tested on farms in the UK and Portugal, with trials planned in Australia.
🧬 Biotechnology
AlphaProteo generates novel proteins for biology and health research
DeepMind released AlphaProteo, a new AI system that designs new proteins. Researchers at DeepMind hope AlphaProteo will help accelerate advancements in drug discovery, disease diagnosis, and biosensor development. The new system has demonstrated impressive success, with 3 to 300 times better binding affinities compared to existing methods across a range of target proteins, including those related to cancer and viral infections.
A newly approved ‘living drug’ could save more cancer patients’ lives
In February 2023, the FDA approved Amtagvi, a TIL (Tumour-Infiltrating Lymphocyte) therapy that extracts immune T cells from a patient's tumour, grows them in a lab, and reinfuses them to attack cancer. Now, the first results from clinical trials have been published, and while not effective for all patients, the therapy shows promise for solid tumours. Scientists are exploring further applications across other cancers.
‘Right to Repair for Your Body’: The Rise of DIY, Pirated Medicine
What do you do when the price of medicine is too high but you have an anarchist and hacker mindset? You make DIY versions of expensive pharmaceuticals at a tiny fraction of the cost and teach people how to make their own versions of expensive drugs at home, circumventing patent laws. This is what Four Thieves Vinegar Collective is doing, offering DIY versions of drugs like misoprostol (abortion medication) and Daraprim (HIV medication). Mixæl Swan Laufer, the spokesperson for the group, criticizes in this article the pharmaceutical industry's high costs and outdated intellectual property laws, aiming to make DIY drug production accessible and common while emphasizing ethical considerations.
Defense-Forward Biosecurity
As the cost of DNA sequencing and analysis decreases, new ideas become possible. One of them is real-time biosurveillance systems to detect emerging pathogens before outbreaks occur. Advances in sequencing technologies and AI are making pathogen detection faster and more scalable, though challenges remain in cost and implementation. These systems could enable early identification of novel diseases, allowing for quicker public health responses and better global coordination to prevent future pandemics.
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!"
Thanks a lot for these summaries and excellent articles. They're really helpful to keep updated.