Blockchain is a decentralized digital ledger used to record transactions and is the underlying technology behind cryptocurrencies such as Bitcoin. However, blockchain technology is now also used in many other areas. It could possibly soon take on a central role in the AI sector.
• Discussions about combining AI and blockchain at the World Economic Forum in Davos
• Blockchain could bring transparency to AI training data
• IBM and Casper Labs are working on a solution
AI models like the chatbot ChatGPT are trained using huge amounts of data. However, this means that the AIs are automatically only as good as their training data – and this can sometimes be permeated by human prejudices or even contain false information. As a consequence, the answers of a generative AI would then also reflect these prejudices or misinformation. Blockchain technology could offer one way to prevent this.
Advertising
Trade Bitcoin and other cryptos via CFD (also with lever)
At Plus500 you can bet on rising and falling crypto prices – even with leverage. Try the free demo account now!
Plus500: Please note the Hints5 about this advertisement.
AI training data book as an important use case of blockchain
As “CNBC” reports, the World Economic Forum (WEF) in Davos, Switzerland, discussed, among other things, how blockchain could be used to prevent biases from being contained in AI training data and thereby shaping the AI’s responses. According to “crypto.com”, a very big problem in the field of artificial intelligence is that the decisions made by an AI are usually not transparent for the user and the data used is not known, which can lead to trust problems. Blockchain could help here. The technology behind Bitcoin & Co. offers a decentralized and transparent method for recording data. Every step in the AI training data lifecycle, from collection to use, could be tracked on the blockchain. This would create an immutable data history that would make it possible to verify the integrity and provenance of the data, thereby also creating more transparency and verifiability in AI systems. According to “CoinGeek”, this would also create a neutral record of which sources were used – and to what extent.
In Davos, Sheila Warren, CEO of the Crypto Council for Innovation, even described a blockchain-based AI training data book as a possible “killer use case” for blockchain technology, according to “CNBC” and was confident that AI and blockchain will work together in the near future will go hand in hand. “I actually believe that the verification of an AI, and in some ways the system of checks and balances within an AI system, will be blockchain-driven and blockchain-supported,” Warren said. And she could be right, because there are already companies working on such a system.
The first solutions are in the works – and software giant IBM is getting involved
“In the product we’re building, the records are actually checkpointed and stored on the blockchain so you have proof of how the AI is trained,” said Medha Prlikar, co-founder of blockchain company Casper Labs, according to “CNBC ” during a panel discussion at the WEF. To achieve this, Casper Labs is working with the software giant IBM. According to the press release, both companies are currently co-developing a solution that will also support version control using blockchain serialization capabilities, allowing companies to efficiently rely on previous iterations of an AI system in the event of performance issues or skewed output. “If you’re using the AI when it’s learning and you notice that the AI starts to hallucinate, you can actually reset the AI. And so you can undo some of what it’s learned and go back to a previous version of the AI,” Prlikar explained according to “CNBC” in Davos. A “hallucination” describes the tendency of some AI models to recognize patterns that are not present in the training data, which negatively affects the quality of the AI’s responses.
“While generative AI is rightly exciting businesses for its transformative potential, its practical applications have been severely limited by the inability to monitor and respond to the AI systems that feed data,” Mrinal Manohar, CEO of Casper Labs, said in the joint press release quoted with IBM. “With IBM’s help, we want to provide not only a better way to understand why AI systems behave the way they do, but also a clearer path to resolving behavior when hallucinations or performance issues occur. The long-term potential of AI will depend on how effectively and efficiently companies can understand, control and respond to increasingly large AI training data sets,” said Manohar.
Blockchain could also help AI with copyright law
But blockchain can not only help improve the explainability of AI. A blockchain-based AI training data ledger would also make it much easier to identify the original “owner” of the data and automate the process of payment, according to “CoinGeek”. Currently, AI companies are repeatedly exposed to copyright lawsuits. For example, the “New York Times” accuses the software companies OpenAI and Microsoft of unlawfully copying and using millions of articles from the “New York Times” to train the chatbot ChatGPT and is demanding billions in damages. In the future, blockchain technology could be used to track exactly which content was actually used as training data – and the rights holders could be automatically compensated.
Editorial team finanzen.net
Selected leverage products on IBM
With knock-outs, speculative investors can participate disproportionately in price movements. Simply select the lever you want and we will show you suitable open-end products on IBM
The leverage must be between 2 and 20
Advertising