Technology

Brilliant Frame, ADeus, Mobile Aloha, Flowpilot — Four Groundbreaking Open-Source And AI Gadget Projects

Karan Kamble

Feb 13, 2024, 06:31 PM | Updated Feb 14, 2024, 10:09 AM IST


Frame Glasses by Brilliant Labs. (picture by Brilliant Labs)
Frame Glasses by Brilliant Labs. (picture by Brilliant Labs)

The last few months have seen a slew of devices being brought to market that are combining artificial intelligence (AI) and new form facts of consumer electronics.

While large companies are still perfecting their products, quite a few open-source, hobby projects are also showing us that open source and cheap electronics can still give large companies a good fight for their money.

‘Open source’ refers to software or projects where the original source code is made freely available and may be redistributed and modified. It allows anyone to see, use, and improve the code — like a recipe that's open to everyone, so people can improve it and share their versions with others. This collaborative approach often leads to high-quality software that is free to use for everyone.

Here are four examples of gadgets or devices that leverage open-source communities and allow you to cook up your own variation.

1. These Brilliant ‘Frame’ AI glasses

Turning a pair of glasses into a powerful AI gadget is genius! Brilliant Labs’ “Frame” is just that. The Singapore-based startup, led by former Apple employee Bobak Tavangar, describes their innovation as “a reinvention of eyewear for the AI era.” Frame was launched on 8 February.

The “multimodal” AI glasses — meaning they can accommodate various types of input and output, like text, visual, and audio — can identify, characterise, and reimagine visual objects, translate text from one language to another, and run an AI-based web search. See the ad for Frame:

The fully open-source wearable tech features augmented reality (AR) optics with an organic light-emitting diode (OLED) display. The visual recognition and analysis is powered by ChatGPT maker OpenAI, translation by OpenAI’s Whisper automatic speech recognition system, and web search by AI startup Perplexity’s conversational search engine. All the elements are tied together by the Noa app after it’s paired with the glasses.

Brilliant Labs invites anyone to “hack, build, and modify,” speaking to the open-source nature of the AI device. Frame is open-source under the MIT licence.

Brilliant’s other product is Monocle, a device meant to be attached to a pair of glasses. It is described as “a pocket-sized AR device for the imaginative hacker.”

2. Adeus Wearable AI Device Of The Humane Kind

Software developer Adam Cohen Hillel has put together a fully wearable, open-source AI wearable device. Adeus captures what you say and hear, and transcribes and stores it on your own server. Thereafter, it can be engaged in a conversation using the app, and it will be able to respond based on what it has picked up so far. 

This wearable tech is a personal AI in the vein of Avi Schiffmann’s Tab, Rewind Pendant, and Humane Ai Pin, but with a stronger emphasis on owning your own data. 

“In the upcoming world, Personal AI will become an integrated part of our daily lives, they will be with us all the time, and will know much about us as our closest friends — and therefore, with Adeus, we are making sure it is completely open-source, and that you can own you own data,” Hillel writes.

What Adeus looks like (Photo: Adam Cohen Hillel)
What Adeus looks like (Photo: Adam Cohen Hillel)

The Adeus hardware constitutes three elements — a mobile or web app, a hardware device, and supabase.

The app is a way for the user to interact with their personal AI and data through chat, the hardware device is the wearable gadget that will record whatever is said or heard and send to the backend for processing, and the Supabase is the backend and database, where the data will be processed and stored, and a large language model (LLM) of your choice will be leveraged.

Hillel says Adeus can be set up “today, not in a few months (oh, and it's cheaper!).” Here’s a demo of the device by the maker himself.

Hillel is also the maker of Deepen, an AI self-care companion that helps with mental well-being.

3. A Humanoid Robot For House Chores

Stanford University and Google DeepMind researchers have developed a robot that can carry out household tasks like cooking, cleaning, and giving you a high-five upon co-training. The robot’s strength lies in its ability to imitate humans.

While such imitation learning isn’t new in robotics, it hasn’t led to useful outcomes in most cases due to limitations in mobility and dexterity. The project team of Zipeng Fu, Tony Z Zhao, and advisor Chelsea Finn have, however, taken imitating learning to a whole new level.

The result, called Mobile Aloha, is a system for imitating mobile manipulation tasks that are bimanual and require whole-body control. Mobile Aloha is a low-cost, whole-body teleoperation system for data collection. “Using data collected with Mobile ALOHA, we then perform supervised behavior cloning…,” the researchers write.

Demonstrating a particular task to the robot 50 times can raise the success rate by up to 90 per cent. Examples of complex mobile manipulation tasks that Mobile Aloha can autonomously complete include some cooking, using a cabinet to store cooking vessels, pushing a chair, calling and entering an elevator, and rinsing a used pan using a kitchen faucet. 

Impressive, isn’t it? See the various demonstrations and training videos for yourself. And while you’re at it, see these robot fails, for fun:

Mobile Aloha moves as quickly as a human, at 1.42 metres per second, and can carry 100 kg of weight. The robot can reach 65 cm to 200 cm vertically, and 100 cm away from its base.

What’s more, all its hardware and software is open-sourced.

4. Putting A Car On Autopilot Using A Mobile Phone

Imagine downloading software on to your mobile phone that then drives your car for you — this is possible, thanks to Flowpilot.

Flowpilot is an open-sourced advanced driving assistance system based on openpilot that can give your car autopilot functionality. (Openpilot can drive for hours without driver action, and works on over 250 car models of Toyota, Hyundai, Honda, and others.) It is able to carry out functions like adaptive cruise control, automated lane centering, forward collision warning, lane departure warning, and driver monitoring.

Flowpilot runs on Linux, Windows, and Android-powered machines like smartphones. The software was developed by Mankaran Singh, who works at Ola Electric and runs FlowDrive.AI on the side, building his advanced driving assistance system along with other engineers.

Singh published a video on X in May 2023 which drew the interest of many (990.2K views as of writing this article) — it shows him using his second-hand Xiaomi Redmi Note 9 Pro running Flowpilot to drive his Alto K10 car.

FlowDrive has an industry-ready solution for the market, for which they are seeking startup collaborations.

Read here to know how to run Flowpilot on your car.

Karan Kamble writes on science and technology. He occasionally wears the hat of a video anchor for Swarajya's online video programmes.


Get Swarajya in your inbox.


Magazine


image
States