Blockchain has the potential to mitigate bias in the datasets utilized for training artificial intelligence models, presenting a promising application for this technology, as per insights shared by executives with CNBC.
One of the primary apprehensions regarding AI models, such as those supporting applications like ChatGPT, is the presence of biases or inaccuracies within the training data. This situation could lead to the AI system providing responses that incorporate these biases and erroneous information.
The inception of blockchain in 2009 with the introduction of bitcoin brought forth a revolutionary concept. In the realm of bitcoin, blockchain serves as an unalterable and secure public ledger for transactions. Businesses are now exploring the adaptation of these principles to various other blockchain applications, also known as distributed ledger technology.
In the context of AI, the training data can be stored on the blockchain, enabling the developers to monitor the data used for training the model effectively.
Casper Labs, a blockchain company focusing on enterprise solutions, recently joined forces with IBM to establish such a system.
Medha Parlika, the Chief Technology Officer and co-founder of Casper Labs, highlighted this advancement during a panel discussion at the World Economic Forum in Davos. She explained, “The datasets in our product are checkpointed and recorded on the blockchain, providing a verifiable record of the AI training process. In case the AI displays signs of generating inaccurate information, developers can revert the AI to a prior state, undoing some of the learning progress.”
The term “hallucinations” is used to describe instances where an AI system produces false outputs.
Blockchain technology has been a topic of discussion across various industries for years, with sectors like finance and healthcare exploring its potential applications.
Sheila Warren, CEO of the Crypto Council for Innovation, expressed her belief that a blockchain-powered ledger for AI training data could serve as a groundbreaking application for this technology. Warren stated, “I firmly believe that the validation and oversight mechanisms within an AI system will rely on blockchain technology for support and verification.”
