One of the long-running jokes between my husband and me involves how robotic developments have happened at such a glacial pace that Rosie the Robot from The Jetsons is still futuristic. I say to him, where’s my robot cook? Where’s my robot housekeeper? Not to mention that flying cars remain a pipe dream because AI is still incapable of effectively driving earthbound ones.
However, if you believe Google’s latest, its Gemini robotics division has made great strides in how robots can react in real time. In recent video releases we’re seeing how its robotic arms can shift and adapt as verbal commands change. When researchers moved containers around to interfere with the robot’s ability to follow commands, the arms adapted immediately, putting objects in the right containers anyway. That may sound simple, but it’s quite a breakthrough. AI is making this adaptability possible.
A Chinese company has been touting labor-saving humanoid robots that can fold clothes, prepare food (from partially prepped ingredients), vacuum and do many other household activities. A video of a robot doing chores shows it working so slowly that a human can outwork it in a quarter of the time. To me its only viable use would be for people who turn it on in the morning and come home after it’s finished. It also could be quite helpful for disabled or elderly people who have difficulty handling these activities by themselves. The only problem for average people is the price.
The Shanghai Kepler Robot Company is rolling models off its assembly lines priced at 1 million yuan, the equivalent of US$137, 775. That’s far more affordable than anything currently available in this country, but still far too costly for use beyond corporations and the extremely wealthy, and it’ll probably be cheaper to hire people for the tasks that they can do anyway.
A Norwegian startup called 1X plans to test its Neo Gamma robot in a few hundred thousand homes this year as part of an effort to train them for daily activities. But get this: 1X teleoperators will monitor the robots remotely as part of the program, and be able to take over their activities when they fail to work right. The resident can decide when to turn the robot on and give company employees access.
The Sunnyvale, California company Figure is building humanoid robots for factories and warehouses, and plans to move into the home market. Founder Brett Adcock says in a video that household robots are exponentially harder to train than those doing repetitive tasks in factories.
One of the most exciting announcements came in March, when Nvidia revealed its alliance with Disney Research and Google DeepMind to create an open-source robot infrastructure. It will be powered by a next-generation physics engine called Newton, which will guide the robot’s movements. Its AI foundational model, Groot N1, will power the planned humanoid robot’s ‘thinking.’
At a recent Nvidia AI conference for developers, CEO Jensen Huang introduced Blue, an entertainment robot that uses Newton and is similar to Star Wars characters like R2-D2 — it’s child-sized and responds to commands with beeps and whistles. Disney will use it in its parks.
Boston Dynamics, which has been entertaining YouTube fans with its dancing-humanoid-robot videos for years, has the potential to enter the home arena, but for now makes its money with DARPA and NASA contracts. Its sole public-marketing product is its security robot dog, which is currently on guard duty at Mar-a-Lago. But as technology refinements make it cost-effective to produce, I wouldn’t be surprised to see it enter the consumer market on a wider scale.
Open AI touts GPT writing skills, faces copyright suits
Recently OpenAI CEO Sam Altman bragged that his company’s latest ChatGPT version can do creative writing extremely well. Considering that the company fed it practically everything written in English, it had some great writing teachers. However, it also scraped everything it could off the internet, in addition to many libraries’ worth of books and magazines. With OpenAI now fighting a flurry of lawsuits over its use of copyrighted content, Altman is crying in court that the company has to be able to use the material (without paying for it) to compete in the global AI realm.
Anthropic, which was launched by founders who left Open AI to build a more ethical model, built the LLM called Claude with sourced content. However, now in its third year, it’s gaining a reputation for throwing caution to the wind to be more competitive, scraping websites with reckless abandon in pursuing content to train Claude.
The only LLM that is at the point of hiding its plagiarism is Grok, because it’s capable of taking written materials and summarizing them in its own unique way through general artificial intelligence.
Regardless of what US courts decide, European courts will likely have a lot to say about these pirating models.
As a writer, I am playing the world’s tiniest violin in my head over their legal trouble for ripping off the hard work of writers trying to make a living. Money will pour into these AI companies, and the least they can do is compensate living authors whose ideas are being stolen — even if it’s over time.
