News
“Peregrine Mission” from Astrobotics, carrying tiny Mexican robots, burns over the ocean: It was supposed to be the first US mission to the moon since the Apollo missions and the first moon mission for the Mexican Space Agency. But having launched last week (8 January) from Cape Canaveral, it suffered a leak, and the command center decided to make it return to Earth and burn in the atmosphere, possibly throwing debris into the ocean. The mission’s payload included several tiny robots developed in Mexico under the “Colmena” project.
Google wrote a ‘Robot Constitution’: On JanJanuary 24oogle’s DeepMind released a set of rules for its AI products called “Robot Constitution” (got to concede the name is catchy). Despite its title, it’s not a written “Constitution” but “a set of safety-focused prompts” to an LLM that instructs this one to avoid dangerous tasks or things such as sharp objects or electric hazardous appliances. The announcement has been received with a range of reactions from existential (“make sure its new AI droids won’t kill us”) to bland (“make sure AI robots behave when tidying your home”).
OpenAI’s custom GPT Store is now open: Yes, you can offer your instance of ChatGPT with customized capabilities to the public –but not yet charge the users money for its use. Available GPTs range from help to “design presentations or social posts with Canva” to “search and synthesize results from 200M academic papers.” In my view, GPTs are mini-apps that specialize in a concrete use case and help users with specific needs. This is part of the evolution of Generative AI systems I talked about in the article “How Generative AI Will Disrupt Everything During the Current Decade,” going from a do-it-all chatbot like ChatGPT to very specialized products developed by small teams –or even individuals.
This week’s product
The most surprising new hardware product is the Rabbit r1. It's essentially a small, handheld gadget that aims to help the user by acting as an app-free intermediary to the actual apps.
The punch, though, doesn’t come from the device itself, which is a low-end plastic handheld red portable device (see figure on top), but from its AI muscle. The Rabbit can control your apps in order to achieve your goals and execute its actions through those other applications.
So, in a way, it’s an app controller that understands your requests by using an LLM that they call “Large Action Model” (LAM) because the output is actions instead of just words. I view the LAM acronym more like a publicity gimmick than a paradigm shift, but the way Rabbit uses third-party apps is innovative indeed: they use machine learning and lots of videos about using those apps in order to learn how to control them. That makes sense because, in the mentioned videos, there is audio as well, explaining the intention of interactions, and this information can be useful for training the AI.
Another innovative aspect of Rabbit is that it explicitly avoids offering apps, as they consider apps a bad UX paradigm. I consider apps bad, too, and later on, I’ll perhaps write a blog post explaining this.
A final groundbreaking aspect is the Rabbit’s price: only $199 dollars! At this price, it’s appealing to give it a shot. Many have done, and Rabbit’s initial set of units has just been sold out.
What is…?
Cooling cities using Deep Space radiation.
Yes, as it sounds. And this is not a hyper-futuristic laser-beaming to be used in a couple of centuries, but something to be deployed right now.
We all know that most forms of energy (including fossil fuel) on Earth are derived from the solar light that bathes us. But in places with heavy insolation, the problem is not getting sun but getting relief from the excessive heat.
It’s the coldness of deep space that can be used to lower the temperature of rooftops.
The idea is to set up surfaces that can radiate heat to the sky and traverse the atmosphere, getting away from Earth. To radiate heat to the sky would be great because, for comparison, domestic air conditioning units expel the heat to the surrounding air, making the overall city heating problem only worse. In contrast, radiated heat would get out of the planet.
Or would it?
In reality, things are not as simple as just radiating heat upwards because the atmosphere is not transparent to any infrared radiation. It’s been shown that only the infrared in the band of 8- to 13-micrometer-wavelength can traverse the atmosphere and escape the Earth altogether.
So the challenge is to find ways of producing emitters of infrared radiation exactly in that band, which is not easy to do.
Several researchers have proposed different materials for infrared radiation in the right band, including glass, polymers, and more. There is even a small company, SkyCool, that is trying to commercialize deep space heat radiation.
On a global level, if deep space heat radiation was used at scale, “134 Million metric tons of carbon would be reduced,” according to SkyCool website.
Compared to using solar panels for electricity production, “a SkyCool panel saves 2x -3x as much energy as a solar panel generates given the same area.”
Isn’t this technology cool?
What I’m up to
I examined the alternative administrative procedures for patenting my “Silent Voice” technology, and I found out that the best option goes as follows:
First, file a national patent in any country affiliated with the Paris Convention. It’s going to be in Mexico.
In less than 12 months from there, file an international “PCT” patent.
If needed, after 20 months from the beginning, file an application in each individual country where protection is wanted.
This means I have to file a national patent ASAP, which is simpler and cheaper than an international patent.
By the way, I can help you out with a free consultation if you have questions related to patents –no hidden catch at all. You know, previous to “Silent Voice,” I patented –and obtained the patent certificate for– an electronic intelligent pillbox for my university.
For the consultation, just book an appointment on my web calendar at: https://rbrena.youcanbook.me/
Blog piece highlights
This time my Medium article “User Manuals Are Being Replaced By Chatbots” discusses one application of AI that has not been heavily hyped, which is kind of puzzling, given how popular AI has become, and also how deeply we hate user manuals.
The highlights of this story are:
User manuals suck, to the point that many people skip them. However, many of them end up underusing their devices.
It would be a lot better to have specialized chatbots to answer our questions directly instead of pointing us to a manual page.
In my 2024 AI predictions, I mentioned that user manuals will be replaced by interactive chatbots.
The low reliability of LLMs is being corrected in restricted settings using the “RAG” technology.
Volkswagen was the first major car company to bring this interactive help system to new vehicles.
Gradually, all car brands –as well as all sorts of companies– will jump on this train, given the advantages of interactive help, to the point that user manuals will go the way of the Dodo.
Here is the end of the free newsletter. Below I’ll give a “friend link” to the blog piece, as well as friend links to some curated good articles I found on Medium.