Proposal: Dall-e for the physical world

Proposals Round 6
#1

Name of Project: Chiral

Proposal in one sentence: Chiral is a new AI model that can create physical parts, structures, and entire machines from a description in natural language.

Description of the project and what problem it is solving:

The limitations of traditional engineering are determined by human imagination and the tools at our disposal. Engineering tools haven’t changed much in 500 years. Drawings are still made on screens, even with CAD software. Designing complex 3D objects using only a flat sheet of paper is very difficult. We need tools to express our thoughts. Our natural way of thinking does not scale with CAD tools. Scaling our engineering output requires a different approach.

Our platform allows engineers to encode laborious human-driven engineering processes into the AI model. Then with a text description, the AI can create objects that are more complex, more functional and more customized to their use cases. The created objects are ideally suited for mass-production using 3D printing.

To name a few, climate change, zero hunger, and access to clean water are all fundamentally engineering challenges. We need to move engineering towards an exponential paradigm as soon as possible in order to solve these challenges.

Grant Deliverables:

  • An open-source implementation of text-to-3D for the community to build on.
  • An exploratory user interface for Chiral, aimed to be accesible to everyone from curious to domain-experts.
  • Optional: 3D Parametric Mesh Dataset used for train Chiral.

Spread the Love

Happy to hire and partner with Algovera members for this initiative. Special interest if your experience is related with advanced User Interfaces, UX/UI experience is the most important part on new AI products (not the model) and machine learning practitioners with experience on 3D deep learning or similar.

Squad

Chiral is my personal side-project, and actually going solo founding it.

  • Twitter: @arthur_hyper88
  • Arthur_88#3168
  • 0x91DAb5930e60BE490d6512ee1cb984ad7b5088a8

deck-2
deck-21574×865 158 KB

3 Likes
#2

Review:

Any proposal that puts a new, powerful, open-source tool in the hands of developers must be a good thing and should yield - as the author says - a paradigm where problems only have to be solved once before being available worldwide.

A couple of questions would be whether the model could be used to design and mass-produce firearms etc and whether there are any safeguards against this and also transparency about any commercial use of the community additions.

2 Likes
#3

That is a good point. At the moment the dataset used has no weapons. But the debate is much deeper, for example if the model designs a sword or a tank, should it be blocked or altered? What kind of responsibility does the AI model have towards the user and how can we safeguard the best possible balance. It is definitely an interesting ethical discussion, but it is too early to have it.

#4

Great project, especially for climate change.

Do you have a focus in mind or specific scenarios?

If you implement smart green cities for example, you could integrate sustainability factors such as green roofs, solar panels and IoT data etc.

Do you also have some effort in mind for these grant deliverables and how many hours you could work on this? Perhaps focus on some smaller items per feature.

It would be great to create some wireframes too.

Grant Deliverables:

An open-source implementation of text-to-3D for the community to build on.
An exploratory user interface for Chiral, aimed to be accesible to everyone from curious to domain-experts.
Optional: 3D Parametric Mesh Dataset used for train Chiral.

Good luck and enjoy!

#5

Of course, it is not about solving all of humanity’s problems, but about giving the right tools so that everyone can come up with solutions regardless of their experience, background, education or location. The idea of Chiral is to stimulate the imagination, not to replace the process. A tool that makes us think and challenges our knowledge will always be very useful. We are seeing how stable diffusion models suddenly allow millions of people to transform their ideas into drawings, and in the process are challenging design styles or art in the process. Chiral aims to do the same for industrial design.

Certainly if I get this grant, it will serve as validation and this will become my full time job and I hope to develop the idea to its full potential. I have no time limit for it.