Volume AMA: AI & Decentralized GPU Access in Blockchain

This week’s AMA episode explores decentralized GPU access as an area where the Paloma Chain can shine and impact the blockchain AI space. Taariq Lewis @LewisTaariq, CEO of VolumeFi (


) shares compelling use cases where Paloma validators can help reduce barriers to GPU access for users and data scientists. Paloma can automate faster, cheaper, and smoother payments for decentralized GPU access available on-demand and on-chain, execute transactions on behalf of users who use model training and model inference while ensuring privacy, and enable users to reap the benefits of decentralized GPU farms. Conversations with potential partners are in the brew as Paloma gears up to join the blockchain AI landscape.

Paloma is relaying again! The network is faster and enhanced with all the new powerful features of Cosmos SDK 50. The Palomabots for TWAP, leverage, Juice, and limit orders are back online. Curve’s Llamalend bots are prepped for launching.

To jumpstart relaying activity, there are zero fees on Volume for the whole month of May. Users will only need to pay minimum fees for gas making Palomabots the cheapest bots to trade, TWAP, and leverage trade.

Taariq welcomes everyone to Volume’s Twitter Spaces.

As the DeFi market slows down and weaves through high interest rates, Paloma is looking into Blockchain and AI particularly, decentralized GPU access to ensure a strong, hardcore launch for the GRAIN token. Taariq shares the Volume’s goals as they shift their focus to Blockchain and AI space.

He welcomes all 32 validators to the Flock. Tumbler, the new chain, is cranking and relays are fast and working on Paloma. One of the goals is to find more avenues of profits for Paloma validators, and the AI blockchain space is the most promising.

Taariq congratulates all validators for the massive success of the recovery of the 12ETH that was trapped on the old Messenger when the network was going through a major upgrade.

The team is finalizing the bots to allow the users to close their bots and withdraw their 12ETH.

He shares that Paloma has started relaying again! All relay fees are paid back by Volume to the validators.

The work on Pigeon Feed will restart on May 20th, and the launch is expected in late June.

Taariq assures that the GRAIN token will have a price when Pigeon Feed goes live. It will require a DEX and a peer on each DEX. It will likely be Curve for Paloma on the DEXs and


The focus is to try to get more demand for bots. With the DeFi bull run slowing down considerably, Paloma is launching more new bots and exploring the blockchain and AI space.

Volume has zeroed out the fees on all the bots to jumpstart relaying activity. Taariq stresses that all service fees that Volume is charging are now zero. Paloma will only be charging minimum fees for gas.

Paloma’s TWAP bots are back online and going for intent plays like CoW Swap. Limit Order bots are also coming back soon.

Paloma is Curve Llamalend bots in addition to the Curve Leverage bots which are now back online.

Taariq notes there will be continuing weakness in the market since Bitcoin dropped to 61 and ETH to 3000. There will be less trading volume and more farming.

Squeeze, Paloma’s EVM L2 will be launched after the bots go live. Squeeze points will be given to users and can be converted into Paloma GRAINS. In the future, this will give users opportunities to farm.

All these events will be leading to the launch of the GRAIN token, notes Taariq as Pigeon Feed nears completion.

While it is still too early to discover the demand for decentralized GPU access, Taariq cites the rise of Akash’s trading range since its token launch in 2021. From sub $1, the Akash token has risen between $3-$6 in the last three months alone. The price is a reflection of people’s enthusiasm for the space. Taariq believes that there is something valuable there, and it is possibly AI.

He talks about the mechanics of Akash and how they rent out GPUs which must be done with the Akash token or another intermediate token. Users cannot use cash. There might be an opportunity for Paloma to be involved with integrating a more seamless, simpler payment system specifically for users who want to pay in dollars or any other token in a decentralized way.

He expounds that users who want GPU access from Akash can deposit funds to make a payment to rent a GPU and Paloma validators would execute the swap automatically.

Palomabots will make payments fast and easy for GPU access. With faster access to GPUs, more people will use them.

Taariq explains that model training is essentially the art of updating the parameters on an AI model such that it is now tuned to a particular use case.

He notes that this process takes a lot of GPU power and GPU time. A model that has Llama for example has about 70 billion parameters will require GPUs to process a big mountain of data and it will take time. This is what GPU farms are good for.

Model inference, on the other hand, is about giving a very complex question to the model and seeing what the returns are. This also takes time as GPUs crank through data, a vector of words, and the relationship between words and inference.

Taariq notes that with model inference, models get smarter the more feedback they receive. Users will want to use a cash flow, in a way that they do not want to overpay.

He believes in providing access to more of these models to train and to do inference cheaply and fast. It is what’s going to make it really impressive for the decentralized model providers.

There is stiff competition in the data center space and providers of GPU access are going to make it faster, easier, and cheaper for people.

Privacy is another way that Paloma validators can contribute especially for users who want to keep their business private.

Users can give Palomabot instructions about where their data is residing. They can put in how long they want to train for and how much they want to use it. Validators can simply sign a transaction to move it and get it onto a server. Payment is taken care of.

The Palomabot can be programmed to do all these activities in a way that maintains privacy. Users stay disconnected from their data so they can train whenever they need to, get the results, and not be tied to the user personally.

Paloma will be able to handle payments, data routing, retrieval of results, and how to run another model again.

Taariq assures that the team will continue to experiment and have conversations with teams in blockchain and AI. They will start with integrating easy and fast payments so that more people can use these services without pain.

Taariq reminds everyone that Paloma’s relaying execution is live.

Plans are in motion to invite blockchain GPU access players to talk about their services, what they are looking for, their pain points, and most importantly, how Paloma can align with their vision.

Pigeon Feed will come online in June and deployed in DEXs where there will be price discovery on the GRAIN token. They will continue to scale and enable validators to collect revenue from users.

Stay tuned for the next AMA!

To find out more about Volume, check out Volume Finance (


), join the Discord (

), and follow us on Twitter (@Volumefi). Check out Paloma Chain on




), follow them on Twitter (@palomachain), and join the flock on Discord (

). Coo! Coo!