Welcome to Cognitoid network

Artificial General Intelligence algorithm on its own Layer 1 blockchain

Join Network

The real AGI

Able to extract semantics from the environment like human brain does, not just statistics.

Multimodal

Modality agnostic, supporting any input modalities like text, visual, audio etc.

Decentralized

Highly parallel nature makes it as strong as big the underlying network is.

Hierarchical

Can aggregate traditional neural networks for efficient preprocessing of inputs.

About Cognitoid

This is a very brief overview of the algorithm. Detailed paper will be made available soon.

The Cognitoid algorithm performs uniform generalization and prediction of raw environment. It extracts semantics and causality from the input data rather then averaging out statistics like traditional neural networks do. Unlike traditional neural networks Cognitoid doesn't require millions training passes and is capable of one shot learning. With subsequent fine-tuned updates and growing network Cognitoid has potential to surpass human mind. The algorithm is conceptually very strong but the implementation is experimental and is still at its early stage. A lot of updates is coming. The client node has built-in update feature and will inform you upon updates available so you don't need to do anything to stay synchronized.

View Whitepaper   View Algorithm Paper

The algorithm is highly parallel and its power is fully determined by the size of network it runs on. Albeit it is possible to run it on a single computer (i.e. sandbox mode) or even on a supercomputer it won't be able to perform at highest power this way because processing would take too long. On the contrary, the speed of individual nodes is not so important. You don't need a high end CPU or graphics card to run it, an average PC will be just fine. You do need a stable internet connection though.

  • Neurons are distributed between active nodes, partially overlapped for stability.
  • Intelligent units ("cognitoids") are trained individually. It is possible to fork them to share knowledge.
  • Neural consensus is used to prevent output biasing or other manipulations.

Network and Blockchain

Cognitoid network is powered by a Layer 1 blockchain that ensures stability, security and optimal load balancing.

Green

No energy-consuming mining is involved. PoW requires nodes to be online and process neural signals, that's all.

Lightweight

There are no smart contracts or non-native tokens, just pure tokenomics to power the AGI. It's like Bitcoin of the AGI age.

Fair Mining

250 AGI tokens are currently mined every block (60 seconds) and equally distributed among all active nodes.

Earn Passively

750 AGI tokens are currently distributed to staking every block (60 seconds).

Earn By Training

If you train a great intelligent unit ("cognitoid") you can charge other people for querying it as well as other cognitoids forked from it.

Trading Available Soon

Peg-tokens are planned to be launched on major popular chains as well as creating liquidities.

Setup Node

Cognitoid v0.5.3 is available for Windows. Other platforms support is coming soon.

Once launched the node will download the blockchain (it will take some time) and start earning AGI tokens for you!

Download Cognitoid

Frequently Asked Questions

If you have more questions feel free to ask in Official Community.

  • Is this really an AGI?

    The algorithm in single-node sandbox mode is already capable of extracting semantics and doing one-shot learning in simple arithmetical and logical tasks. Now it's time to expand the network and start fine tuning it for much more complex and impressive things. We are at the start of the journey and the way ahead is now well seen. And yes, this is an AGI! Weak neural networks can't extract semantics from any environment and make their own conclusions. They just average out known examples they've learned.

  • When the network is big enough and its parameters are tuned accordingly it should potentially surpass human mind. It can help solving hardest problems of science and greatly improve everyday life of people. But of course there is also a countless number of simpler use cases (human-like chat bots, resource management agents, robotic avatars, AGI expert units in specific domains like law, maths, marketing etc)

  • No information about this algorithm was ever published before. The first paper providing an overview of Cognitoid has been made available for the public in June 2023.

  • Some parts of the algorithm are exponentially slow when run on a single machine. Even thousands of CPUs won't completely solve this. A superior intelligence ideally requires a global network with dozens millions nodes. That is the goal and the reason for creating the network.

  • Yes! While it's critically important for the algorithm to have a lot of nodes in the network it's not so important for every single node to be extra powerful. You just need an average PC to run a node. But please make sure your internet connection is stable, that is important.

  • You can earn passively by running a node. The more developed Cognitoid ecosystem becomes the more you will earn. But you can also train some unique intelligent unit ("cognitoid") and earn much more from those who query or fork it.

  • Cognitoid token (AGI) is the native token of Cognitoid blockchain.

  • 3,000,000,000 AGI was premined. The algorithm is not yet fully parameterized so a large part is reserved for experimental cognitoid training and fine tuning the network. This will end up being paid to the network and distributed as profits to active node owners. Besides that, after successful launch some tokens will be sold to setup liquidities on major public blockchains (Ethereum, BSC, Polygon). Peg-token emission will be limited to keep the price higher.

  • No. Cognitoid's main focus is Artificial Intelligence so it doesn't make sense to overload the network with irrelevant features. Those looking for advanced financial functionality can use Ethereum or a plenty of other blockchains supporting that. Cognitoid token (AGI) will be soon available there via peg-tokens.