Giving artificial intelligence models consent to use content for training is a “perfect use case” for blockchain technology, according to Avery Ching, co-founder and chief technology officer of Aptos.
Speaking at India Blockchain Week 2024, Ching emphasized how blockchain could enhance control over how AI models are trained.
He highlighted the potential for blockchain to provide clear consent mechanisms for determining whether specific content can be used for AI training.
How blockchain can improve controls over AI training data
Ching said that AI and blockchain are the “buzzwords of the day” in many spaces. However, he stressed that despite the hype around large language models (LLMs), what he finds interesting is the data used to train AI models.
Aptos chief technology officer and co-founder Avery Ching at the India Blockchain Week Conference 2024. Source: Cointelegraph
Using the event as an example, Ching noted how content from the stage — recorded on video or written in news articles — could be used to train AI. He explained:
“There’s going to be a question on whether we want that content to be allowed and whether we want that content to be trained on. So, this is a perfect use case for blockchains because they’re immutable. They provide a center of control.”
The Aptos co-founder said with blockchain there are ways to control whether people allow AI to access their content. “As a natural back end for this kind of information, it’s perfect,” Ching added.
Related: Musk again asks to block OpenAI’s ‘illegal’ conversion to for-profit model
Potential application on social media platforms
Ching, who previously worked at Meta, acknowledged the scalability challenges of applying such controls to platforms like Facebook, where billions of pieces of content are generated daily. Adding blockchain-based permissions for AI training to each piece of data could involve significant transaction throughput and costs:
“If you’re gonna add controls around each of those pieces of data, there’s gonna be a high transaction throughput, there’s gonna be costs involved.”
Despite the challenges, Ching said Aptos is actively working on supporting this type of use case within its blockchain network. He sees blockchain as a natural back end for controlling how content is used, particularly as AI continues to evolve.
Magazine: 5 incredible use cases for Based Agents and Near’s AI Assistant
Giving artificial intelligence models consent to use content for training is a “perfect use case” for blockchain technology, according to Avery Ching, co-founder and chief technology officer of Aptos.
Speaking at India Blockchain Week 2024, Ching emphasized how blockchain could enhance control over how AI models are trained.
He highlighted the potential for blockchain to provide clear consent mechanisms for determining whether specific content can be used for AI training.
How blockchain can improve controls over AI training data
Ching said that AI and blockchain are the “buzzwords of the day” in many spaces. However, he stressed that despite the hype around large language models (LLMs), the thing he finds interesting revolves around the data used to train AI models.
Aptos chief technology officer and co-founder Avery Ching at the India Blockchain Week Conference 2024. Source: Cointelegraph
Using the event as an example, Ching noted how content from the stage — recorded on video or written in news articles — could be used to train AI. He explained:
“There’s going to be a question on whether we want that content to be allowed and whether we want that content to be trained on. So, this is a perfect use case for blockchains because they’re immutable. They provide a center of control.”
The Aptos co-founder said with blockchain, there could be a way to control whether people allow AI to access their content or not. “As a natural back end for this kind of information, it’s perfect,” Ching added.
Related: Musk again asks to block OpenAI’s ‘illegal’ conversion to for-profit model
Potential application on social media platforms
Ching, who previously worked at Meta, also said platforms like Facebook generate billions of unique content per day. While blockchain can help with AI training consent, the executive recognized that this may involve high costs. He said:
“If you’re gonna add controls around each of those pieces of data, there’s gonna be a high transaction throughput, there’s gonna be costs involved.”
Despite this, the executive said that this type of use case is what they are working to support as they build the Aptos network.
Magazine: 5 incredible use cases for Based Agents and Near’s AI Assistant
