Bonsai

Why Bonsai?

The next generation of truly intelligent machines will, just like humans, need to learn. We build the tools to teach them and the operating system to power them.

Bonsai is driven by a simple but far-reaching question: how does one program a machine that is truly intelligent to do something useful? You certainly won't program it the way we do now. Simply put, if it is in fact truly intelligent, then by design it must learn, which is to say, it must be taught. This then leads to the conclusion that we must have tools that enable us to automate instruction so that the intelligent machine can be programmed.

We have taken this insight, explored it extensively, and crafted a new foundation for building intelligent systems.

What is Bonsai?

Bonsai AI does for artificial intelligence what databases did for data.

Where databases provide the basis for managing data, our technology provides the basis for managing models of intelligence. This encompasses a new programming language (analogous to SQL), platform (analogous to a database), and the associated development tools. By moving up the level of abstraction, we enable companies to build sophisticated AI systems using their existing programmers.

From a nuts and bolts perspective, this means a new high level declarative programming language (named Inkling) in which developers codify the concepts associated with an area of knowledge they'd like the system to learn, and a curriculum for doing so. This is then compiled into a mental model (a specialized set of graphs) which the runtime uses to construct an appropriate learning topology (e.g. neural networks) and to direct and assess training. The low level details of the learning algorithms and hardware are handled transparently by the platform.

How?

Not just your basic, average, everyday, ordinary, run-of-the-mill, ho-hum artificial intelligence company

We blend rigorous and sophisticated engineering with cutting edge neuroscience. Each informs the other, and together they enable us to push forward the state of the art. Yet we are not a research operation. We are building a very real platform, language, and tool chain to solve problems today.

What sets us apart is that our system is designed so that users work at a higher level of abstraction. They thus benefit from the ever evolving learning algorithms and hardware without having to modify their own code.

We are looking for exceptional individuals to join us in building the next big platform, and we look forward to talking to you.