The best insights from Pontem’s panel with Aptos CTO Avery Ching

A couple of weeks ago, we were lucky to have the CTO of Aptos Labs Avery Ching in our Twitter Space with Pontem’s Alejo Pinto. They discussed the challenges the AI poses and how to overcome them using blockchain, We’ve summarized the best insights for you here.

Avery Ching isn’t just a co-founder and CTO of Aptos Labs – here’s also a renowned expert in blockchain and Big Data. In the past, he ran both of these departments at Meta itself as it was working on Diem, the blockchain project that gave rise to Aptos. That’s also where Avery met Mo Shaikh, the current Aptos CEO.

Ching joined our recent X Space with the Chief Growth Officer of Pontem Network, Alejo Pinto. The conversation revolved around AI – a timely topic for us as we are preparing to release our own crypto chatbot, PontemAI.

Aptos is well-suited for AI-scale content verification

For starters, Alejo described the two biggest issues that AI and especially Large Language Models (LLMs) like ChatGPT have created for us all as content consumers:

1) It’s harder and harder to distinguish between honest and reliable content from stuff created using AI with the purpose of deceiving and confusing people (“fake news”);

2) When you use an LLM, you provide lots of data to the corporation that owns it, and this makes AI a centralizing power.

However, Alejo Pinto believes that a solution exists, and it’s blockchain technology. First of all, you can use it to verify who created a piece of content, when, etc.; and second of all, users can decide which data to share with the LLM owners.

To this Avery Ching said that blockchain can indeed help track content provenance. Example:  a politician orders for AI-generated content to be sent out, claiming that an opponent had expressed some objectionable views. Blockchain verification could help readers understand who ordered these allegations to be disseminated and if they were true or not.

At the same time, Ching made it clear that blockchain infrastructure can’t yet handle such quantities of data, given that Meta Alone generates 4 billion pieces of content every day. Aptos does have what it takes, though: it’s supremely scalable and fast, so it could become the base for such a verification system. However, said Avery Ching, blockchain teams shouldn’t focus too much on competing with each other, since the real competition is between Web3 services and centralized providers.

Why decentralized and centralized platforms should work together

Speaking of centralized vs. decentralized, Alejo Pinto said that regular users will choose decentralized services only if their UX and UI are at the same level with centralized APPs – which, sadly , is almost never the case.

Avery Ching agreed, saying that most users care about the costs and the speed: they like their apps “cheap and fast”, not about privacy. We can’t expect them to change their mind, but we can try to make blockchain better in terms of UX.

Another very important insight from Avery Ching: decentralized services should be prepared to collaborate with centralized cloud providers whenever it ensures important benefits in terms of cost, data availability, etc.

“Web3 can’t be good for everything. It has serious cost trade offs. In some cases these tradeoffs will be worth it, such as when you need to know if a piece of content was generated by a human or an AI, when, and why. In other cases, you can work with the cheaper and faster centralized infrastructure. The two can augment each other and compete at the same time.”

Alejo then touched upon the second big use case for blockchain in AI: user data control. With blockchain, users can decide which information can be used by the company that provides the AI – and even get rewards for it. Avery Ching said:

“If your user data goes into training an AI model, you should be able to opt out of it. The same goes for the content we create on social media etc. Blockchain changes how we think about creating content and authorizing its use in the context of AI models.”

Blockchains have to be ready for higher loads

Apart from infrastructure development, there is another challenge that Web3 platforms have to be ready for as AI if they want to host AI apps: a large number of transactions. A spike in transaction volume can cause an outage.

According to Avery Ching, it’s all about having robust preventive practices in place. When Alejo Pinto asked him about a pause in transactions that happened on Aptos after a code change a few weeks ago, Avery answered:

“Outages happen even to the biggest centralized services like Meta. I’d say that Aptos has better preventive practices than those guys: we use not one, but two reviewers for every code commit, and we conduct massive sets of tests before every update. In this case, we found the bug in the code very quickly and fixed it. We didn’t lose a single user transaction – they were all eventually confirmed. This shows that in terms of security procedures, Aptos is ready for the arrival of AI applications.”

AI comes with a host of challenges: ethical concerns, content attribution, fake news, and others. Here at Pontem, we believe that blockchain can help solve all these, but it will take serious improvements in infrastructure. We’ll keep covering AI in our blog articles and panels – plus, stay tuned for the release of our crypto chatbot, PontemAI!

Listen to the full version of the podcast on Pontem Network’s Twitter page.