3 Key Points
- MIT developed an AI-powered kitchen device called Kitchen Cosmo
- Uses a camera to recognize ingredients and prints customized recipes
- Introduces the concept of Large Language Objects that extends LLMs into the physical world
What is Going On?
MIT architecture students have developed an AI-based kitchen device called Kitchen Cosmo.[MIT News] Standing about 45cm (18 inches) tall, this device uses a webcam to recognize ingredients, accepts user input through a dial, and uses a built-in thermal printer to output recipes.
The project was conducted at MIT Design Intelligence Lab, led by Professor Marcelo Coelho. Graduate student Jacob Payne and design major Ayah Mahmoud participated in the development.[MIT News]
Why Is It Important?
Honestly, what makes this project interesting is more about the philosophy than the technology itself. Professor Coelho calls it Large Language Objects (LLOs). It is a concept of taking LLMs off the screen and moving them into physical objects.
Professor Coelho said, “This new form of intelligence is powerful, but it remains ignorant of the world outside of language.” Kitchen Cosmo bridges that gap.
Personally, I think this shows the future of AI interfaces. Instead of touching and typing on screens, you show objects and turn dials. This is especially useful in situations where your hands are busy, like cooking.
What Will Happen in the Future?
The research team plans to provide real-time cooking tips and collaborative features for multiple users in the next version. They also plan to add role-sharing functionality during cooking.[MIT News] Student Jacob Payne said, “AI can help find creative ways when figuring out what to make with leftover ingredients.”
It is unclear whether this research will lead to a commercial product. However, attempts to extend LLMs into physical interfaces will increase in the future.
Frequently Asked Questions (FAQ)
Q: What ingredients can Kitchen Cosmo recognize?
A: It uses a Vision Language Model to recognize ingredients captured by the camera. It can identify common food items like fruits, vegetables, and meat, and generate recipes considering basic seasonings and condiments typically found at home. However, specific recognition accuracy has not been disclosed.
Q: What factors are reflected in recipe generation?
A: Users can input meal type, cooking techniques, available time, mood, dietary restrictions, and number of servings. They can also select flavor profiles and regional cooking styles (e.g., Korean, Italian). All these conditions are combined to generate customized recipes.
Q: Can the general public purchase it?
A: Currently at the prototype stage in MIT research lab, no commercialization plans have been announced. Since it started as an academic research project, commercialization is expected to take time. However, similar concept products may emerge from other companies.
If you found this article useful, subscribe to AI Digester.
References
- Counter intelligence – MIT News (2026-02-03)