Fusing blockchain with AI to fight bias and deepfakes

In the last two years, artificial intelligence (AI) has been a rapidly evolving field. It has found applications in various domains, from healthcare to consumer goods, and has contributed to significant technological advancements and breakthroughs. However, it doesn’t come without its challenges.

Most issues revolve around AI’s ability to proliferate the creation and spread of misinformation and disinformation. Others revolve around the creation and spread of biased information. But recently, blockchain technology has emerged as a potential solution to the problems that AI systems experience.

Unlocking AI’s black box

Currently, much of AI’s decision-making process is a ‘black box’ to users and even some developers. The opacity of these AI models leads to uncertainty and distrust, as people are unclear about the rationale or even the legitimacy behind the outputs they receive. Even the creators of these AI models warn their users that their AI models can make mistakes, which is why the user should double-check important information.

However, a blockchain could be used to log, monitor, and reference the data that AI models are trained with to serve as a transparent source of truth. A system like this would timestamp the instances where an AI model was given new data and trained so that if or when an AI model produced a biased output, the operator would be able to analyze the ledger to understand when the model began to produce those sorts of outputs, and which dataset may have caused the model to react that way.

Using a blockchain while training an AI model would create a chronological, unalterable proof of how the AI was trained, ensuring that any changes, updates, or deviations can be accurately tracked and analyzed. This level of documentation is crucial for auditing AI systems, primarily when they are used in critical applications that require high levels of accountability and transparency.

Tackling the Deepfake dilemma: Blockchain as a verification tool

Because generative AI systems are now capable of producing high-quality images, audio, and videos, they can easily be used to deceive the public. The creation of this fake multimedia, especially when it involves the impersonation of people, is often called deepfakes; deepfakes often portray individuals in situations or actions that never occurred, causing misinformation and reputational harm.

As generative AI products continue to improve, the realism of these deepfakes improves as well, and we are now at a point where deepfakes are often indistinguishable from genuine content at first glance. At the moment, a few solutions are being used to determine whether a piece of multimedia is a deepfake. One solution relies on third parties and individuals close to the source to verify the legitimacy of the content. However, this method is not always timely. Other solutions involve companies embedding watermarks or invisible markers in AI-generated content. These markers can be visually identified or detected in the codebase, signaling whether the content is authentic or fabricated—but not all companies are doing this yet. However, blockchain can also be a promising solution to these problems.

By creating a blockchain repository of verified and original content, any multimedia asset not present in this repository could be considered questionable or unofficial. This is just one way a blockchain-based system could be used to combat the deepfake problem. For instance, a politician and their team could use such a blockchain-based system so that the public can independently determine the authenticity of their published content. Anything the politician or their team has not recorded on their blockchain log could then be treated with skepticism.

Implementing a blockchain repository for multimedia assets presents a proactive approach to address the deepfake challenge. It provides a transparent, immutable record of genuine content, which helps distinguish authentic materials from deepfakes. While it’s not a foolproof solution, it could be a step in the right direction in the fight against the deepfake phenomenon as it promotes authenticity and trust in digital media.

Maximizing AI transparency and combating bias:

Using a blockchain with AI systems presents a promising solution to enhance transparency and fight bias. To fully realize the benefits of these innovations, there will need to be a concerted effort from AI developers, users, and regulatory bodies to establish common standards and practices for implementing blockchain in AI systems.

It is also important to note that not every blockchain is a good candidate to be implemented into the type of AI/blockchain system described in this article. The systems described above require large amounts of data to be sent to various locations at frequent intervals, sometimes several times daily. If a blockchain has low transaction throughput or high transaction fees, it will most likely be technologically constrained or inefficient to run from a cost perspective for it to make sense to use that blockchain as a solution to these problems that AI experiences.

Regardless, the fusion of AI and blockchain can potentially solve some of the growing problems in the artificial intelligence space. Bringing these two technologies together can quickly identify when and why bias occurs in an AI system and can combat the rising issue of deepfakes. There are most likely individuals experimenting with this type of blockchain AI solution at the moment, but for this type of system to be mass-adopted, it will require a collective effort from technologists, policymakers, and users. Educational campaigns directed toward AI users and the public about how blockchain enhances AI systems and how it can increase trust in AI would be needed. If that were to take place, both technologies would experience higher acceptance rates as more individuals could be confident in the systems that they are using.

In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI. 

Watch: Artificial intelligence needs blockchain