How are AIs and Prediction Markets Related?

Predictions aren’t perfect. AIs can’t make them 100% perfect either. Still, with the advent of platforms like Augur and Gnosis in the last few years, an interesting question has come to light. Can AI systems play a key role in scaling and improving prediction markets?

More specifically, can AIs make the outcomes of prediction markets more accurate?

Before we answer that question, it’s important to define what a prediction market is. In short, it’s a place where anyone can bet on the outcome of any event. These outcomes are then verified by some sort of impartial team, or in the case of blockchain-based prediction markets, properly incentivized reporters. The problem with all of this, in a basic sense, is that everything in such a system can go wrong due to human error.

Arguably, the only way to avoid this is to involve computers in the reporting or verification process. To understand the possible advantages and disadvantages of such a move, it’s important to zero in one specific projects from the blockchain space.

For our purposes, we’ll stick with Augur and Gnosis, which are, for now, the most well-known blockchain protocols that are designed to support decentralized prediction markets. Both allow anyone to build a market where a prediction is made and crypto is staked on it. In Gnosis, however, the team claims to involve AI in the reporting process while the Augur team makes it clear that they are confident with humans every step of the way.

Before you assume, however, that the Augur developers have completely discounted the role of AI and ML in streamlining the reporting process, keep in mind that they do use decentralized oracles to verify events.

Fully explaining exactly how this is the case when their “oracles” are synonymous with “reporters” would call for another dedicated post on the subject. Expanding upon what I suggested above, a network like Augur depends on a group of humans to verify the outcomes of all predicted events and believes they are decentralized from the prediction process. Still, it is easy to argue that this is not a perfect solution. Humans are humans and humans are fallible.

Even so, both Augur and Gnosis believe that prediction markets can serve a greater purpose. This is perhaps, best said in the Gnosis whitepaper in which the team claims that their protocol will essentially underpin at least part of the Internet of Things going forward.

For that to happen, however, the entire process of prediction and verification would have to be sped up and made more efficient. It is at this point that AI, or at least, ML, comes in.

If the human element was largely taken out of the reporting process, then it would likely be much faster than an average of 3 days, which is what Augur reports on its’ documentation page. 3 days is far too much of a turnover time for a use case like the Internet of Things which would need to process scores of transactions in a matter of moments.

As we have mentioned before, this is because of the fact that the IoT is a place for machines(including computers) to transact and talk with each other to help each other improve on a technical level.

In essence, if it were to finally scale across every aspect of our lives, it would likely usher in a new iteration of our global economy. For a preview of what that might be, take a look at our post on Fetch.ai here.

Overall, for the IoT to scale in a way that will support our smart devices across the globe, it will have to be able to process transactions more quickly than anything else that currently exists. If we then circle back to prediction markets with this under consideration, then we can add that for them to play a major role in this scaling, they will have to be “scaled.”

At this point, it would also be logical to question why prediction markets are an ideal solution for the IoT? Why not consider some network like Fetch to be the best answer?

The easiest answer to both of these questions is that more than one sort of project can help to support the IoT. No rule is written down that, for example, the IoT can only be dependent on blockchains. What is likely to happen in the future is that all sorts of networks will come together and connect with the help of certain “bridge-like” protocols to make one powerful internet for machines.

For prediction market networks like Augur and Gnosis to play a part in this future, they could improve their efforts with AI systems. One obvious use case would be in the reporting phase, as suggested above.

The teams could make it so smart contracts triggered the usage of an AI reporter at the close of a prediction or “market.” Once this AI reporter was told to take action, it would then verify that the real-life events involved with the closed markets concluded in the way that the human bettors said they did. Imagine a network in which humans can bet on anything and wait only a matter of minutes for the results of the bet to be verified.

Honing in on exactly what type of AI-system would be needed to do this would take a wide-range of technical expertise in both the AI and ML fields. That means that the development teams would have to pay experts in these fields to construct these systems, which would likely not be cheap.

Another, more domain-specific argument against the usage of AI in prediction markets is that smart contracts might be able to do the same for the verification process over time. The reasoning behind this boils down to the utility of what are popularly called “decentralized oracles.”

Generally, “decentralized oracles” are entities that do not expressly belong to the network they are connected to, but play a key role in the verification of its’ transactions. Overall, it’s easy to understand what types of things can be oracles.

According to multiple sources, this includes: stock tickers, sensor data from all sort of things like vehicles, weather instruments, and smart devices, news reports, and much more. Basically, connections are made between one or more of these items and the smart contracts that run the predictions that the network hosts.

When a predicted event comes to past, a signal is sent to the smart contract that tells the decentralized oracle to verify the outcome of the event in question. If, for example, this event is an election, than the oracle would likely comb through multiple news sources to verify the results.

While implementing one or more AI systems across all of a prediction market’s predictions would likely be very costly, it is possible that using certain Machine Learning algorithms could be more viable.

For blockchain networks, a good starting point might be to utilize a suite of automated machine learning software, though in doing so, it would be necessary to choose an option that is open-source. From there, it’s important to note that AML cannot do everything that an ML professional or data scientist can do, at least not yet. There simply isn’t enough time and man power to keep up with the progress that AI and ML are making and put all of the popular algorithmic frameworks in one place.

Before we get lost in the weeds here, let’s pause for a moment. I honestly got so deep into this research that I realized that there are a multitude of connections between AI, ML, and any sort of prediction. For that reason, we’ll save some of this discussion for another time and jump ahead to a final possibility to consider.

If ML algorithms were combined with protocols like Augur and Gnosis, as well as, a decentralized supercomputer like Golem, then in a sense, prediction markets could serve as the foundation for a truly global and reliable IoT.

Keep in mind that all of these industries are early-stage as of yet. As time goes on, it will be interesting to see what other possibilities arise related to prediction markets and the future of all of our technologies. For now, keep in mind that with the right combinations, the possibilities seem to be almost endless.