Using Blockchain to Ensure Responsible Use of Artificial Intelligence | Image:Unsplash
Blockchain technology may have found another critical use case, especially as adoption continues across different sectors. Speaking during the World Economic Forum Annual Meeting earlier in January, industry executives highlighted blockchain’s potential use in artificial intelligence (AI).
Blockchain technology is an adaptable tool usable across nearly all sectors. Even in gambling and online casinos, some of the best sites in the industry have adopted blockchain technology by accepting cryptocurrency payments and letting players use digital assets to place wagers and withdraw earnings. This option makes iGaming platforms more inclusive because players can enjoy exciting casino games regardless of geographical restrictions.
In addition to inclusivity, blockchain contributes significantly to security, ensuring that all transactions are immutable and can be audited at any time. For example, in the renewable energy sector, blockchain payments facilitate secure and transparent transactions for buying and selling energy, allowing consumers and producers to directly engage with each other, which promotes the use of sustainable energy sources and contributes to a greener environment.
But what about when it comes to AI? Can blockchain really help ensure this growing technology is used more responsibly?
Blockchain to Help Reduce AI ‘Hallucination’
According to stakeholders, the immutability of blockchain technology can be the ‘killer use case’ for AI, as Crypto Council for Innovation CEO, Sheila Warren suggested. One major problem with artificial intelligence and large language models (LLMs) is the possibility that they may produce misleading information. In some cases, AI chatbots may provide false or slightly biased responses that could mislead users. Stakeholders are now suggesting that blockchain technology can help to solve this problem.
The responses AI tools provide largely depend on their training. These bots are trained using data that could, unfortunately, contain some inaccurate information. To solve this problem, AI firms can use blockchain technology to track the training process for reference. According to Casper Labs co-founder and Chief Technology Officer Medha Parlikar, the company is developing a product that creates checkpoints for the datasets used for AI trading. Parlikar explains that this helps to “roll back the AI” and undo some of the learning, or return to a previous accurate version if the AI begins to hallucinate. Generally, a hallucination refers to AI models producing biased, misleading, or incorrect responses.
The Intersection of Blockchain Technology and Artificial Intelligence
Since the popularity of artificial intelligence exploded after OpenAI launched ChatGPT, there have been several submissions about how blockchain and AI can integrate to improve data usage and availability. In addition to using blockchain technology to create a verified source of data for AI training, the technology can also help with automation. Since the technology is decentralized, all of the computing power and general operations required to power the AI tool can be automated, reducing the level of human supervision necessary for training.
The decentralisation in blockchain technology also simplifies AI training and maintenance. Ordinarily, these tasks consume a lot of resources for storage, hardware, software, and general maintenance. However, blockchain simplifies these requirements because the technology is distributed across thousands of nodes. This means that the pressure of maintaining an artificial intelligence tool or network is shared, making operations easier and more efficient.
One major criticism against artificial intelligence is the seeming lack of privacy. Companies launching AI projects train these chatbots or services with all kinds of data regardless of sensitivity or privacy concerns. While some AI chatbots try to limit the amount of sensitive data presented to users, the risk of privacy breaches still exists. Blockchain can help create immutable records of potential privacy risks, such that all existing and future training data can be audited using these data points, to properly protect data and reduce potential risks.