Hello and welcome to Weekly News Roundup Issue #479!

This week, Figure unveiled their second-generation humanoid robot, Figure 02, featuring a sleeker design and upgraded AI.

In other news, OpenAI delivered more drama with two key employees leaving the company, and advised not to expect GPT-5 to be revealed during this year’s Dev Day. Meanwhile, Groq raised $640 million to meet demand for fast AI inference, and video game actors are on strike over AI concerns.

In robotics, DeepMind trained an off-the-shelf industrial robot to play table tennis, and another company trained a robot dog to carry groceries.

Enjoy!

Figure unveils a new humanoid robot

Figure AI has unveiled their second generation humanoid robot, Figure 02, which bring many improvements over its precedesor, Figure 01.

The new robot has a new, sleeker body hiding cables and motors under plastic matte black plastic panels. It also has upgraded hands capable of carrying up to 25 kg (55.1 pounds) and designed to manipulate objects that a human hand can grasp. A new swappable battery with 50% greater capacity allows the Figure 02 robot to run for five hours on a single charge, according to the company.

Figure 02 has also become smarter. The robot now has three times more computing power and is equipped with an onboard vision language model (VLM). Using six cameras, it can perceive its surroundings and perform tasks autonomously. Thanks to a partnership with OpenAI, Figure 02 can understand natural language commands and engage in conversations with humans. Although Figure did not show these capabilities for Figure 02, they showed how it works with Figure 01 a couple of months ago.

However, Figure is just one of many new players on the humanoid robotics scene. For long time, the only notable company building humanoid robots was Boston Dynamics which both amazed and terrified millions of people who watched on YouTube what Atlas can do. But in the last 2-3 years, an entire cohort of new humanoid robotics companies emerged seemingly out of nowhere all at the same time. There are big companies like Tesla which is building Optimus and promises to start using it in its factories next year. And there are a bunch of new companies. Figure is one of them. Other include Agility Robotics, whose robot Digit has recently landed its first official job, 1X, Apptronik, Sanctuary AI, Unitree, and more.

I admire the engineering that went into building these machines. They are a remarkable achievement, combining mechanical engineering, material science, electronics, sensor technology, software engineering, and machine learning. But the question I have is: will humanoid robots take off?

All those companies promise effectively the same thing—a replacement of human labour with robotic labour. A drop-in replacement of human workers without any significant changes to the workplace. These robots are supposed to come in, take the tools humans use and start working after some basic training. They can work all the time, they don’t complain, and are cheaper and more scalable than human labour. At least that is what is promised.

Some of these robots are already being tested in real-life workplaces. Agility Robotics’ Digit is being trialled at Amazon, while the Figure 01 robot is being tested by BMW. Mercedes is trialling Apptronik’s Apollo for "low skill, repetitive" tasks. Sanctuary AI has partnered with European carmaker Magna to deliver their Phoenix robots to Magna's factories. Another car company, Dongfeng Liuzhou Motor, has also started to incorporate humanoid robots in their factory.

However, these are still pilot programmes. The real test for these robots will be their interaction in actual workplaces with real customers, who expect them to deliver value rather than just serve as novelties. Overpromising is one of the biggest risks facing humanoid robotics companies. The flashy demos we see on YouTube might give a false impression of the true capabilities of humanoid robots.

The contact with reality will be harsh, and I believe the first wave of humanoid robots will not meet the high expectations set for them. Many humanoid robotics companies may not survive this initial reality check. The critical question will then be what lessons the surviving companies learn and transform their humanoid robots into a product that the market wants.

Will humanoid robots take off? Conrad Gray · Jun 11 Read full story

Figure is a very well funded startup. To date, the company raised $854 million from Microsoft, OpenAI, Amazon, Nvidia, Intel Capital, and more. After the last Series B funding round in February 2024, Figure was valued at $2.6 billion, making it the most valued humanoid robotics company. Figure looks like will have enough runway to develop 03, maybe 04 and 05 robots, giving them more time to find the right market fit for their robots. Other companies may not be as fortunate.

If you enjoy this post, please click the ❤️ button or share it.

Share

Do you like my work? Consider becoming a paying subscriber to support it

Become a paid subscriber

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

The Race to 150
This article discusses the concept of radically extending the human lifespan to 150 years and beyond. It proposes achieving this through technologies that replace damaged tissues, organs, or entire systems to maintain youthfulness and functionality, as well as through gene therapies, cellular reprogramming, and advanced drug delivery systems to repair and maintain the body at a cellular level. The article also analyses current progress and bottlenecks, calling for a more aggressive approach to achieving radical life extension, framing it as a critical choice for humanity's future.

Neuralink has implanted second trial patient with brain chip, Elon Musk says
Neuralink has successfully implanted its device in a second patient, designed to enable paralysed individuals to use digital devices through thought alone, according to Elon Musk. Neuralink is testing the device, which aims to assist people with spinal cord injuries. The first patient has been able to play video games, browse the internet, post on social media, and move a cursor on his laptop. However, a month after the first implant was installed, 85% of its flexible threads, which record neural activity, retracted from the patient’s brain. Time will tell if the second implant performs better.

Brain-Computer Interfaces: The Promise of Technological Telepathy and More Conrad Gray · Jan 29 Read full story

🧠 Artificial Intelligence

OpenAI co-founder Schulman leaves for Anthropic, Brockman takes extended leave
Two more high-profile OpenAI employees have left the company. John Schulman, one of OpenAI's co-founders, has departed to join the rival AI startup Anthropic. Peter Deng, a product manager who joined OpenAI last year after leading products at Meta, Uber, and Airtable, also exited some time ago, the company confirmed. With Schulman's departure, only three of OpenAI's 11 original founders remain: CEO Sam Altman, Greg Brockman, and Wojciech Zaremba, lead of language and code generation. Meanwhile, OpenAI Greg Brockman is taking a sabbatical until the end of the year to relax.

OpenAI tempers expectations with less bombastic, GPT-5-less DevDay this fall
We should not expect GPT-5 to be revealed during this year’s OpenAI DevDay, according to the company. Instead, DevDay (or DevDays, as there will be sessions in San Francisco, London, and Singapore) will focus on educating developers about what OpenAI has to offer and how to include OpenAI models in various products and services.

OpenAI won’t watermark ChatGPT text because its users could get caught
According to The Wall Street Journal, OpenAI has developed a watermarking system for ChatGPT text and a detection tool. The method alters word prediction patterns to create a signature without affecting text quality and is 99.9% tamper-resistant. However, OpenAI is holding back its release due to potential business impacts and concerns from users, especially non-native speakers. OpenAI is exploring alternatives like embedding metadata, but it is still early in the process.

A New Type of Neural Network Is More Interpretable
Kolmogorov-Arnold Networks (KANs) are a new approach to neural networks that can surpass traditional models used in vision algorithms, image generators, and chatbots. They offer greater flexibility in data representation while using fewer parameters. Additionally, as this article explains, KANs are easier to tweak, interpret, and understand compared to conventional neural networks.

Ampere AmpereOne Aurora 512 Core AI CPU Announced
Ampere has unveiled a new 512-core Ampere AmpereOne Aurora processor designed specifically for cloud-native AI computing tasks. The company says it will have three times the performance per rack of current AmpereOne processors and will feature integrated AI acceleration and HBM memory. Additionally, Ampere markets the upcoming chip for both training and inference.

Groq Raises $640M To Meet Soaring Demand for Fast AI Inference
Groq, a startup that builds fast chips for AI inference, has raised $640 million in a Series D round, bringing its valuation to $2.8 billion. “You can’t power AI without inference compute. We intend to make the resources available so that anyone can create cutting-edge AI products, not just the largest tech companies,” said Jonathan Ross, CEO and Founder of Groq.

Video game actors are officially on strike over AI
Members of the Screen Actors Guild (SAG-AFTRA) have begun striking against the video game industry after negotiations for a new contract, particularly around AI protections, failed. The union demands safeguards for voice and movement performers against AI-generated content, but video game companies offered limited protections. The strike, which started on 26th July, affects over 160,000 members and impacts projects from major publishers to indie studios.

The State of Chinese AI
Air Street Press published an excellent article exploring the Chinese AI scene. The analysis covers the state of Chinese AI research and industry, comparing it to Western competitors in terms of capabilities and the unique challenges faced by Chinese AI companies, such as complying with CCP political censorship. The article also examines the impact of US sanctions and how Chinese AI companies have adapted to them.

Replacing my Right Hand with AI
Erik Schluntz, an engineer at Anthropic, recently broke his arm and used voice-to-text along with Claude, an AI tool, to continue coding. After two months, he found the experience so effective that he declared he was “never going back.” Schluntz believes AI-powered coding tools could revolutionize the role of engineers, shifting their focus to high-level decision-making and creativity while AI manages repetitive tasks. He also sees AI tools lowering the barriers to software development, enabling more people to bring their ideas to life.

If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?

Refer a friend

🤖 Robotics

Ukraine's new secret weapon: Silent Chinese robo-dogs in urban combat
Robots are finding their way onto modern battlefields, including the popular Unitree Go2 robot dogs. Unlike drones, these Chinese robots are virtually silent, making them highly useful for reconnaissance missions.

▶️ Pull To Walk Robot Dog Grocery Shopping (0:46)

Robot dogs are cool, but what useful things can you do with them? Well, you can turn your robot dog into a robot mule that carries groceries for you, as shown in this video.

Achieving Human Level Competitive Robot Table Tennis

Researchers from DeepMind took an off-the-shelf industrial robotic arm and trained it to play table tennis. The robot and the AI controlling it became skilled enough to compete at an amateur human level. While an expert human player can still defeat the bot, the system showcases the potential for machines to master complex physical tasks that require split-second decision-making and adaptability. The paper describing the project can be found here.

💡Tangents

▶️ The Physics of Dyson Spheres (18:22)

The idea of Dyson Spheres was a radical proposal by physicist Freeman Dyson—an enormous shell of material enveloping a star. Although Dyson's idea is over half a century old, interest in searching for such objects has only grown. But how would these structures work? Are they even physically possible? And what might they be used for? Prof. David Kipping from Cool Worlds answers these and more questions in this video.

Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.

Share

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!"

Message Conrad Gray