AI Is Having Its Moment, But Is It A Good One?
Could this have been the most consequential week for AI so far?
Aaron’s Thoughts On The Week
What makes us human, I think, is an ability to ask questions, a consequence of our sophisticated spoken language. - Jane Goodall
Raise your hand if your inbox and/or social media feed was nothing but news about AI. Every day this week, there was another announcement about AI. Some commentators were wondering out loud if this week would go down as one of the most consequential weeks for AI so far. It may be.
On Monday, Covariant, an OpenAI spinoff, announced its new RFM-1, a platform like a large language model but for robot language. RFM-1 has been trained on years of data collected from the company's small fleet of item-picking robots used in warehouses around the world. It has also been trained using online videos and text data. In the coming months, Covariant plans to release the model to its customers, which could really show what RFM-1 is capable of. The company hopes that the model will continue to improve its efficiency and capabilities as it is used in real-world applications.
“The vision of RFM-1 is to power the billions of robots to come,” Covariant CEO Peter Chen says. “We at Covariant have already deployed lots of robots at warehouses with success. But that is not the limit of where we want to get to. We really want to power robots in manufacturing, food processing, recycling, agriculture, the service industry and even into people’s homes.”
Quickly on the heels of Covariant’s announcement, Figure a startup that partnered with OpenAI to develop a humanoid robot, released a new video which further made everyone wonder how good is AI getting. , a startup that partnered with OpenAI to develop a humanoid robot, released a new video that further made everyone wonder how good AI is getting.
The full video from Figure is our Video of the Week (see below). It shows a human interacting with a robot named Figure 01. The robot has a natural-sounding conversation with the human, identifying the objects in the environment when asked. The human then asks for something to eat, and Figure 01 hands them an apple. Next, the human introduces some trash and asks the robot to pick it up while explaining why it did so. The robot processes the instruction and executes the command without any issues.
Additionally, the robot explains why it gave the apple to the human, and then demonstrates its dexterity and precision by putting the dishes in the drying rack. The video is embedded below and is worth watching to get a better understanding of the interaction.
It was really a “WOW!” moment and something that blew my mind. Good thing we still have control over their power source.
The Regulators Have Entered The Chat
Then Wednesday came with news that the first wave of AI regulation would be coming. European Union lawmakers gave their final approval to the Artificial Intelligence Act, which will become effective later this year. The act, which was proposed five years ago, has received overwhelming support from the lawmakers in the European Parliament. The AI Act is expected to serve as a global guide for other governments looking to regulate the rapidly growing technology.
Dragos Tudorache, a Romanian lawmaker who co-led the Parliament negotiations on the draft law, stated that "The AI Act has nudged the future of AI in a human-centric direction, where humans are in control of the technology. The technology helps us leverage new discoveries, economic growth, societal progress, and unlock human potential."
Big tech companies have generally supported the need to regulate AI. However, they have lobbied to ensure that any rules work in their favor. OpenAI CEO Sam Altman caused a stir last year when he suggested that ChatGPT maker could pull out of Europe if it can't comply with the AI Act. However, he later backtracked and stated that there were no plans to leave.
The AI Act takes a "risk-based approach" to products or services that use artificial intelligence. Low-risk AI systems like content recommendation systems or spam filters face less scrutiny, while high-risk uses such as medical devices or critical infrastructure will face tougher requirements. Some AI uses are banned, like social scoring systems, predictive policing, and emotion recognition systems in schools and workplaces. Police scanning faces in public using AI-powered remote "biometric identification" systems will be banned except for serious crimes like kidnapping or terrorism. I am still curious about what and how it will be enforced, especially with fines of up to 35 million euros ($38 million), or 7% of a company’s global revenue for violators.
The EU has added provisions for generative AI models (something somewhat new since work started on the law back in 2019), and developers of general-purpose AI models will have to follow EU copyright law. The biggest and most powerful AI models posing systemic risks will be under extra scrutiny. Companies providing these systems must assess and mitigate the risks, report any serious incidents, implement cybersecurity measures, and disclose energy usage. Again, how this will be enforced will be very interesting that the EU is allowing each member country to set up its own AI watchdog, where citizens can file a complaint if they think they’ve been the victim of a violation of the rules. Brussels will create an AI Office to oversee general-purpose AI systems. With so many conspiracy theories out there, I am wonder if the EU is opening up the floodgates to unfounded complaints by people who think AI is behind the scenes doing something to them.
The AI Act will become law by May or June after approval from EU member countries. The rules will take effect in stages. Countries will ban prohibited AI systems six months after the rules enter the lawbooks, while rules for general-purpose AI systems will apply a year after. By mid-2026, all regulations will be enforced, including requirements for high-risk systems.
What Is Next Then?
More than likely, this post will be out of date by the time it is published because the space is growing so fast. I am very pro-AI, but do see the need for standards and regulations. So I can see the EU and other government stances when it comes to AI. US President Joe Biden signed an executive order on AI, with lawmakers in seven states working on their own AI legislation. China's President Xi Jinping proposed the Global AI Governance Initiative, while other countries and organizations, ranging from Brazil to the United Nations, are drawing up AI regulations.
What, though, is concerning is that those who truly understand AI are not helping the regulators. So we get regulations like the EU AI Act that have more questions than answers. Even though, OpenAI CEO Sam Altman walked back his statement on exiting the EU market due to the new law, there are plenty of European companies that have not ruled out an exit.
One criticism of the robot standards-making community is that it takes too long to develop a standard, especially given how fast robot technology is advancing. This is one reason that the newest version of ISO 10218 is getting one of the biggest updates in its history. Since its last update in 2011 (yes, 13 years ago), a lot of new types of robots have come to the market.
With AI, updates and breakthroughs will come much faster than in robotics. We need to build an AI standards-making community fast. If we do not, the regulators will do it for us. The problem is that AI cuts across numerous industries, so there will need to be numerous and diverse AI standards to develop. That is going to be hard, but it is still doable if those in the industry put AI standards creation at the top of their priority list. If anything, by shifting resources away from advancing AI before setting baseline standards to move forward, we can start to calm some concerns in the general public and not get to a place where we have to slam the breaks on everything to do what we should have been doing from the get-go.
Maybe that is when we will know how consequential this week truly was.
Robot News Of The Week
ABB opens refitted, state-of-the-art U.S. robotics facility
ABB Ltd. has opened a renovated robotics facility in Auburn Hills, Michigan, which will support the development and production of leading systems in and for the Americas. The facility marks the company's third global robotics factory expansion in three years, emphasizing its commitment to industry leadership in the US. ABB is investing in long-term growth in the US market, which is predicted to experience an 8% compound annual growth rate (CAGR).
Covariant Announces a Universal AI Platform for Robots
Covariant has launched RFM-1, a platform like a large language model but for robot language. The platform is built from data collected from the Covariant’s Brain AI platform deployment. RFM-1 aims to power robots in various industries, including manufacturing, food processing, recycling, agriculture, and the service industry. Covariant's software is currently deployed on industrial robotic arms doing warehouse tasks and it is promising hardware agnosticism.
The E.U. Has Passed the World’s First Comprehensive AI Law
The European Union has approved the Artificial Intelligence Act, which is set to take effect later this year. The act is expected to guide other governments on how to regulate the fast-developing technology.
Reflex Robotics’ wheeled humanoid is here to grab you a snack
At Modex in Atlanta, Reflex Robotics drew a crowd with its impressive robot. It features a torso mounted to a base that dynamically moves its arms and sensors up and down, giving it dexterity to access shelves at different heights. The wheeled base allows it to navigate tight spaces. The system is primarily tele-operated and can be controlled remotely. Co-founder and CEO Ritesh Ragavender compares the interface to a video game and says the robot is "approaching" human-level efficiency.
Robot Research In The News
A quadrupedal robot can do parkour and walk across rubble
ETH doctoral student Nikita Rudin used his parkour experience to push the boundaries of what ANYmal, a quadrupedal robot, could do. Using machine learning, he taught the robot to scale obstacles and perform dynamic maneuvers. ANYmal learned through trial and error and now uses its camera and artificial neural network to determine what kind of obstacle it's facing and performs movements based on its previous training.
Robot climbs offshore wind turbines and paint while underwater
An experimental undersea robot called the Crawfish has been developed by scientists at Germany's Fraunhofer Smart Ocean Technologies research group. It has the ability to climb vertical underwater surfaces and paint them, making it useful for the upkeep of offshore renewable energy platforms. The robot is tethered and made up of two linked components, a BlueROV2 remotely operated vehicle and a crawler unit with four direct-drive elastomer wheels and several tools. It can be lowered into the water by two or three people and steered remotely by a topside operator. The Crawfish can detect damages, cracks, and anti-corrosion paints with its camera and reader head. It can patch damages using a two-part coating material as well.
Robot Workforce Stories Of The Week
Indiana governor signs bill to expand school eligibility, funding use for robotics grants
Governor Eric Holcomb signed a bill into law that expands robotics competitions in schools. The legislation increases schools eligible for robotics grants and allows the funds to purchase robot-building materials. The law also extends funding to nonpublic schools and redefines "eligible team" to include community-based programs. Representative Chuck Goodrich, the bill's author, said this will allow about 75 more robotics teams to compete.
Inside DENSO Robotics' Newest Training Center
DENSO Robotics' Sales, Application, and Training Center in West Chester, Ohio, provides high-quality customer education and support services. The center trains production operators, programmers, and maintenance personnel on the finer points of today's cutting-edge robotic arms, controllers, and applications. It expands DENSO's customer support services, fulfilling the needs of end-users, approved distributors, and system integrators alike. The facility is equipped with everything to maximize the usefulness of DENSO robot arms in multiple industries.
Robot Video Of The Week
Figure 01 can now have full conversations OpenAI models provide high-level visual and language intelligence Figure neural networks deliver fast, low-level, dexterous robot actions Everything in this video is a neural network.
Upcoming Robot Events
Mar. 18-20 Pack Expo East (Philadelphia, PA)
Mar. 25-27 International Conference on Industrial Technology (Bristol, UK)
Apr. 7-10 Haptics Symposium (Los Angeles, CA)
Apr. 14-17 International Conference on Soft Robotics (San Diego, CA)
May 1-2 The Robotics Summit & Expo (Boston, MA)
May 13-17 IEEE-ICRA (Yokohama, Japan)
June 24-27 International Conference on Space Robotics (Luxemborg)
July 2-4 International Workshop on Robot Motion and Control (Poznan, Poland)
July 8-12 American Control Conference (Toronto, Canada)
Oct. 16-17 RoboBusiness (Santa Clara, CA)
Oct. 28-Nov. 1 ASTM Intl. Conference on Advanced Manufacturing (Atlanta, GA)
Nov. 22-24 Humanoids 2024 (Nancy, France)