What is Decentralized AI?

In one of our older pieces, we shared a quote from Walter De Brouwer, that, “AI needs to be decentralized.” With this, you might have found yourself wondering if this is true and if so, what the connection between decentralization and AI really comes down to.

In attempting to answer this question, we could revert back to examples like the Golem Network. With decentralized supercomputing networks, the potential for decentralized AI networks grows due to the fact that the cost of computing has been exponentially reduced.

To recap, accessing Golem only requires purchasing a certain number of GNT, or Golem Network Tokens. The cost of this varies based on the amount of computing power needed, though it can be as low as 15 cents, depending on what task needs to be accomplished. Generally, this may be considered to be significant because of the fact that running an AI network is far from affordable.

According to one particular report on the subject, as early as the Minimum Viable Product stage, costs can already reach $35,000 to $100,000, or more.

Given the fact that decentralized computing networks can be accessed by just about any computer and yet still have the power of the world’s largest supercomputers, it can reasonably be assumed that using such a network is cheaper. At the same time, the chief problem with making such assertions is that the niche of decentralized supercomputing networks like Golem is in its very early stages. Therefore, there is logically a lack of adequate historical data to back these ideas up, over the long term.

At this particular moment in time, the Golem Network appears to represent one of the most viable use cases of the blockchain beyond the financial world. Simultaneously, it has still had trouble gaining a critical mass of users. which almost begs the question: is there any factual basis for the statement that the AI industry “must be decentralized?”

In short, time will tell.

According to a Medium Post by Jesus Rodriguez, who is the chief scientist at an AI firm called Invector Labs, we might begin to understand this if we take into account the rising trend of “Decentralized Artificial Intelligence.” Contextually, this refers to the industry niche that includes all AI projects that run on Blockchains.

In summary, Rodriguez seems to conclude that decentralized AI is synonymous with making the space, “open source.” This, in turn, may be said to greatly lower the cost of talent needed to develop projects, auditing the code of such projects, and perhaps most importantly, lowering the cost of computing. With this, the connection could be drawn out again between the Blockchain and AI in that overall, the heart of the Blockchain is its’ ability to decentralize just about anything.