Are there any decentralized examples of AI systems with blockchain technology?

I think the best example of AI being deployed on the blockchain is SingularityNET. They just had a successful token sell where they sold out of their AGI token which will be able to be used to essentially "pay" for AI tasks to be done for you. Various AI will be put on the network and able to interact and communicate with each other to get various tasks done. There are some great videos online where Dr. Ben Goertzel explains this further. And here is a link to their whitepaper.

Swarm intelligence is the term for systems where relatively simple agents work together to solve a complicated problem in a decentralized fashion.

In general, distributed computing methods are very important for dealing with problems at scale, and many of them embrace decentralization in a deep way. (Given the reality of hardware failure and the massive size of modern datasets relative to individual nodes, the less work is passed through a central bottleneck, the better.)

While there are people interested in doing computation on the blockchain, it seems to me like it's unlikely to be competitive with computation in dedicated clusters (like AWS).

Maybe the term you're looking for is federated learning. Check out OpenMined project, PySyft and Tensorflow Federated libraries.

I'm currently working on a p2p framework to train with neuroevolution, it will have neat, hyperneat, and eshyperneat example experiments. AI developers will be able to fork and add any experiments they want.

It isn't a blockchain per sé but will have a dht for genes, champion nets, peers of course, and genomes. Neuroevolution training can be done in parallel so the population will be distributed evenly among peers, peers that finish early will help slower peers with their evaluations, and all peers will have a check on one of the nets they evaluated by a random peer to prevent malice and all peers will check the champion of each generation. Any peer that evaluates a net will be rewarded a token that is specific to the generation it helped train and will also download a copy of the champion net, later I plan to allow non training clients the ability to view performance of past champs and purchase them from the network, any peer with a token for the genome they purchase will receive part of the payment. This will not be proof of stake, it will be a proof of work where work is evaluating nets, since this needs to be checked by other nets so people don't just post phony fitness for genomes I'll be calling it proof of fitness.

I'm aware of some works that use blockchains in federated learning scenario for accountability, incentivize participants and so on.

Here's an example, BlockFLA: Accountable Federated Learning via Hybrid Blockchain Architecture (2020), but there are probably many similar works around.

Have a look at the paper Blockchain-Based Federated Learning in Medicine (2020), where the blockchain is used as a "federation" server for improving the parameters of local neural networks.