BitDegree Blog Can Blockchain Tech Make AI Smarter?

The internet is undergoing a metamorphosis. The way the web works and the way we interact with it is drastically changing, and for a range of reasons. It all began with blockchain technologies, which promised decentralization and greater individual autonomy (which make up the backbone of Web3). But as of more recently, another tech field has begun to make its presence felt.In the last couple of years, the world has seen huge jumps in the field of AI. It feels as though there is a constant stream of news regarding AI technologies and new AI tools in development. Not only this, but AI has very recently become consumer-ready in a way that we have never seen before. While this technology has existed for a long time in some form or another, and has been used by organizations for many years, it is only in the last few months that accessibility has opened up to the public.Without a doubt, the two AI concepts that appear to be receiving the most attention right now are text generation and image generation services. Both of these have dramatically affected our online behavior, and it is no surprise that these ideas have captured the imaginations of many. Not only is the everyday user excited, but so too are developers, programmers, and investors in other fields.In particular, there is a huge level of interest arising from the blockchain field, possibly because both fields have become highly influential to the way the internet currently functions. Naturally, developers from these areas are wondering what they can learn from each other, as they are both becoming ingrained in the new fabric of the web.While blockchain and AI are not traditionally thought of as companions, there are several important ways blockchain tech could help AI flourish and become smarter. Let’s take a look at why this is, and how something like this could work.A Collaborative ComputerBlockchain technology is great at creating interconnected systems that all function together. Through decentralization, many machines can combine their storage and processing power to work on one common goal. This could be used for AI systems, where instead of there being one centralized server that holds all the relevant data and processing power, there could be a decentralized network of servers and computers that all contribute in some way.There are two major upshots that could come with this.Faster ProcessorsFor starters, global distribution of server power could lead to faster processing, as there would be less strain than if only one server was handling all inputs. This is because the strain would get spread around a wide range of machines and therefore be fractionalized.This should give speedier responses, which would improve an AI’s intelligence in one simple way: it would be able to work out problems and provide solutions faster. Time could get cut down by a huge degree. This benefit would be felt most by graphic designers who need visuals, and researchers who use AI as an assistant.However, improvements to speed will be very much dependent on whether the blockchain being used is optimized enough to handle the traffic it will receive. This means that most current-day blockchains might not be fit for purpose, as they are all prone to congestion, to some extent (which is where there is an inundation of traffic or activity that the blockchain struggles to handle in a reasonable timeframe).That being said, alternative blockchain structures such as DAGs (or Directed Acyclic Graphs) might be suitable. These are networks that can process multiple actions (or transactions) at the same time, whilst only needing a small subset of nodes within its network to do so. This, therefore, makes them faster than traditional blockchains.Public Involvement of DataThe second major benefit of a globally shared AI computer is that people would be able to submit their own data to it. A core requirement of current-day AI is that it needs a tremendously huge dataset before it can make informed actions. That means it needs access to potentially terabytes of content. The more, the better. AI works best as a sponge, soaking up all elements of this world.Currently, data is fed to these machines directly by the companies who make and run them, but via decentralization, people from around the world could feed it! This could help patch up any intellectual blindspots that an AI might have as the core team who create it are bound to forget or be unable to provide data on every single concept and idea. However, a distributed collective of people could easily provide data on practically everything.Of course, if left unchecked, then the AI could be fed the wrong info, which could lead to inaccuracies. Luckily, blockchain technology can help with this, too! In the same way that people can add data to the blockchain, they could also flag and vote on certain pieces and types of data being given to it. For instance, false information and inappropriate content could get flagged by individuals and then disregarded.If the entire dataset is held in a publicly accessible database (which is essentially what most blockchains are, anyway), then people could look through it and vote against certain pieces of information being used by the AI. People could even flag data for copyright, which could help collapse the currently heated debate as to whether AI should be trained on the intellectual property of other people.However, there is one huge limitation to this. There could very easily be disagreements within the community about what exactly should be included in the data and what should not be. This can get especially heated when it comes to historical sources in relation to marginalized groups. AI has been found to have gender and racial biases in the past, and if there are not enough people from different backgrounds contributing to the data of a blockchain-based AI, then those biases could get intensified. This is something that needs to be considered, should this route of AI maintenance ever get employed.Shining a Light Within the Black BoxCurrent-day AI faces a huge problem. We do not know exactly how they make their decisions, or how they come to their conclusions. We have some level of understanding, but nothing complete and full. The reasoning behind an AI’s output is mostly hidden and concealed within its structure. There is a lack of transparency, which is in major contrast to how the blockchain industry functions, where some might say there is an overabundance of transparency.If we are working with a language- or text-based AI, we can ask it to explain how it came to its conclusions, but then that would simply kick the can down the road, as we would then need to ask it to explain how it came to its conclusions about its conclusions, ad infinitum. For this reason, we call AI a “black box”. So much about the way they think is unknown to us right now. One method of potentially illuminating the box is by imbuing AI with blockchain technology. If every decision that an AI makes is tracked and logged within a database, then we can potentially learn more about how they come to their conclusions.However, this only partially solves the problem, as this might not necessarily solve issues of interpretability. Even if decisions can be logged and tracked, it might still be hard to figure out what exactly the AI is doing to get from one idea to the next, but it would at least show a trendline of activity that could be followed and considered by humans.Weirdly enough, this is an issue we have when trying to interpret other humans, too. We can never truly know what somebody thinks and feels as this is private knowledge, but we can still make some reasonable guesses and inferences based on their activity.Perhaps the best way that a blockchain-tracked AI could be beneficial is that it would allow the public to essentially peek behind the curtain and see for themselves what the AI decides on. That way, they can make their own informed decisions, which further means that developers and computer scientists can interpret it themselves, and make their own predictions.These predictions could help the AI itself, provided it is open source. Or, it could help these computer scientists to work on their own independent research and build AI systems of their own from their findings, created in a way that is smarter purely due to the fact that the devs would have been able to reflect on previous iterations.Interoperable AIThis idea of blockchain-tracked decision-making could also create a type of interconnected ecosystem of AI projects and tools. Currently, AI is being viewed as a type of technological arms race, where different companies compete with each other to have the most robust, multi-faceted, and smart machines possible. But wars like this are very much a Web2 concept, and something that many people are eagar to leave in the past.With blockchain technology, not only could we allow people to take a look at how decisions are made, but we could let AI models study each other under the hood. There would, of course, need to be measures in place to ensure that they do not copy each other, and that they are not too influenced by one another, which would essentially poison the well and create a feedback loop where AI starts to make answers based on artificial data.But if this can be avoided, and AI models can be co-trained with each other, with a shared network that allows them to understand how each of them comes to decisions, then they could definitely become smarter. This could work similarly to how humans learn about the world by interacting with others. We do not necessarily copy how other people behave, but we do observe them and then apply our observations to our own decision-making. If AIs are able to do the same by looking through blockchain-tracked actions, then this could help make them smarter and more powerful.However, it relies on AI companies lowering their weapons and allowing each other’s models to communicate w