Pontem and Aptos Labs Leaders on the Future of AI in Web3
In a recent X Space, Pontem’s Chief Growth Officer Alejo Pinto discussed the challenges and the possibilities of a blockchain-AI crossover with Avery Ching, co-founder and CTO of Aptos Labs. Ching is a recognized expert in both BigCompute and blockchain, having led the data infrastructure and blockchain teams at Meta for years.
Pontem Network is a blockchain studio building for Aptos, the most scalable and production-ready L1 network. Pontem is known for its popular Aptos wallet with over 300,000 downloads, the Liquidswap DEX, and is now building an AI-powered chatbot called PontemAI.
“Infrastructure is holding us back”
Much of the conversation centered on the emerging threats of fake and misleading AI-generated content. Pinto pointed out two such challenges:
- It is becoming difficult for users to tell apart bona fide content from AI-generated “fake news”;
- AI is a centralizing technology: by using LLMs, users help corporations to aggregate data about themselves.
Pinto expressed the hope that blockchain can play the opposite role, allowing users to verify the provenance of content and to keep control over their data. Avery Ching agreed, but added that existing blockchains are not yet capable of handling the quantity of content generated today.
“Blockchain can indeed help trace content lineage, and help us understand who wanted that content to be shown. DLT is the only technology that can provide this kind of assurance. However, Meta alone generates 4 billion pieces of content per day – and there’s also TikTok, Instagram, podcasts, email, and other platforms.
Right now, blockchain infrastructure is holding us back. To support content verification on this scale, we’ll need new infra – and Aptos is well suited for it with its scalability and fast finality.”
However, Ching believes it’s not a race between L1s:
“The real competition isn’t between different L1’s but between blockchain and centralized cloud providers. AI itself isn’t better for centralized or decentralized applications; it’s simply a question of where data resides. But it’s true that AI today does benefit closed information systems.”
“Compete, but also augment”
As the discussion turned to centralized and decentralized systems, Alejo Pinto pointed out that Web3 projects lag far behind in terms of UX:
If an action takes five minutes to compute on the blockchain and five seconds on a centralized platform, it’s difficult to get such blockchain verification technology adopted. We here care about decentralization, but most users like their web services cheap and fast.”
Ching believes that instead of trying to compete with cloud providers on UX, blockchain services should partner with them:
“Web3 can’t be good for everything. It has cost tradeoffs. If cloud platforms provide good infrastructure for decentralized services, then we should work with them. For example, content itself doesn’t need to reside on-chain, while hashes do. The two systems can compete and yet augment each other.
Nevertheless, there will be use cases where those cost tradeoffs of Web3 are worth it, like AI content verification and user data control.”
On user data control, Pinto suggested that blockchain gives users sovereignty over their information and lets them decide which AI companies to share it with, and get rewarded for it.
Ching added:
“Where AI model creators use large training sets of user data, users should also be able to opt out from being used for such training. Blockchain makes us think of how we own content and authorize its use, be it social media, email, etc.”
Can blockchains cope with the high load?
Integration with AI services will exponentially increase the scale of blockchain transactions.. Pinto noted that a spike in usage can cause an outage even on the most robust L1. A few weeks ago, Aptos itself experienced a pause in transactions following a code change.
Ching responded:
“Outages happen even to the best centralized services. Our job as programmers is to have great preventive practices in place. Aptos uses two reviewers for every code commit and massive sets of tests for each release. In this particular situation, we quickly pinpointed and fixed the code bug, and not a single transaction was lost. In terms of safety practices, Aptos is ready for the arrival of large AI dApps.”
The development of AI impacts how we think of content creation. Attribution, misinformation,, and ethical concerns could be solved with an immutable ledger that underlies media platforms. Though it will take significant advancements in infrastructure to achieve this, it would be the most crucial use case in the history of blockchain.
The full version of the podcast is available on Pontem Network’s Twitter page.