In a captivating X Space dialogue, Alejo Pinto, Chief Growth Officer at Pontem, and Avery Ching, co-founder and CTO of Aptos Labs, explored the intricate relationship between blockchain technology and artificial intelligence (AI). Ching, a renowned expert in both BigCompute and blockchain, shed light on the challenges and possibilities that arise when these two powerful forces converge.
Pontem Network, a blockchain studio building for Aptos, has gained recognition for its popular Aptos wallet, Liquidswap DEX, and the development of an AI-powered chatbot, PontemAI. Pinto expressed concern over the growing difficulty in distinguishing genuine content from AI-generated "fake news," highlighting the centralizing nature of AI and the potential for corporations to amass vast amounts of user data through large language models (LLMs).
In response, Ching expressed optimism about blockchain's ability to play a counterbalancing role, enabling users to verify the provenance of content and maintain control over their data. However, he acknowledged that existing blockchains struggle to handle the sheer volume of content generated today.
"Blockchain can indeed help trace content lineage and provide insights into who intended for that content to be displayed," Ching explained. "Distributed ledger technology (DLT) is the only technology capable of offering this level of assurance. However, Meta alone generates 4 billion pieces of content daily, and that's not counting TikTok, Instagram, podcasts, email, and other platforms. Blockchain infrastructure is currently holding us back. To support content verification on this scale, we'll need innovative infrastructure, and Aptos, with its scalability and fast finality, is well-positioned to address this challenge."
Ching emphasized that the true competition lies not between L1s (Layer 1 blockchains) but between blockchain and centralized cloud providers. He asserted that AI itself is not inherently better suited for centralized or decentralized applications; rather, the question revolves around where data resides.
As the discussion shifted to centralized and decentralized systems, Pinto highlighted the significant UX (user experience) gap between Web3 projects and their centralized counterparts.
"If an action takes five minutes to compute on the blockchain and five seconds on a centralized platform, it's challenging to convince users to adopt blockchain verification technology," Pinto pointed out. "We in the Web3 space value decentralization, but most users prioritize affordability and speed in their web services."
Ching proposed an alternative approach, suggesting that instead of competing directly with cloud providers on UX, blockchain services should explore partnerships.
"Web3 can't excel in every aspect; it has cost tradeoffs," Ching stated. "If cloud platforms offer robust infrastructure for decentralized services, we should collaborate with them. For instance, content itself doesn't need to be stored on-chain, but hashes do. The two systems can coexist and complement each other."
Nevertheless, Ching acknowledged that certain use cases, such as AI content verification and user data control, justify the cost tradeoffs of Web3.
On the topic of user data control, Pinto emphasized blockchain's ability to empower users with sovereignty over their information, allowing them to choose which AI companies to share their data with and potentially earn rewards for doing so.
Ching concurred, adding that users should have the option to opt out of having their data used for training AI models. He emphasized that blockchain challenges us to rethink our ownership of content and our authorization of its use, encompassing social media, email, and beyond.
The discussion concluded with Pinto addressing concerns about blockchain's ability to handle the high load associated with AI services. He acknowledged that spikes in usage can cause outages even on the most robust L1s, citing Aptos' recent transaction pause following a code change.
Ching responded by highlighting the importance of robust programming practices. He explained that Aptos employs two reviewers for every code commit and conducts extensive testing for each release. In the specific instance of the Aptos outage, the code bug was swiftly identified and rectified, ensuring no transaction loss.
Ching affirmed that Aptos is well-equipped to handle the influx of large AI dApps (decentralized applications) from a safety standpoint.