Original author: MORBID-19

Original text compiled by: Shenchao TechFlow

Hello everyone, it's a new day and another speculative bet. Recently, AI Agents have become a hot topic of discussion, especially aixbt, which has garnered significant attention recently.

But in my view, this trend is completely meaningless.

Let me explain to friends who are not familiar with Bitcoin terminology. Once users bridge assets to the so-called 'Bitcoin Layer 2 Network (Bitcoin L2)', true 'Non-custodial Lending' cannot be achieved.

All 'Bitcoin Bridges' or 'Interoperability/Scaling Layers' introduce new trust assumptions, with few exceptions like the Lightning Network. So when someone claims that Bitcoin L2 is 'Trustless', you can basically assume this is not true. This is also why most new L2 emphasize that they are 'Trust-minimized'.

Although I am not familiar with the Side Protocol, I can almost certainly say that the so-called 'Non-custodial Lending' claim by aixbt is not true, and this judgment is 99% likely to be correct.

However, I do not fully blame aixbt. It is simply acting on instructions: scraping data from the internet and generating seemingly useful tweets.

The problem is that aixbt does not truly understand what it is saying. It cannot judge the truthfulness of information, nor can it validate its hypotheses with experts, let alone question its own logic or reason.

The essence of large language models (LLMs) is merely word predictors. They do not understand the content they output but instead select seemingly correct words based on probabilities.

If I wrote an article in (Encyclopedia Britannica) about 'Hitler conquering Ancient Greece and giving birth to Hellenistic civilization', then for LLMs, this would become a 'fact', a 'historical' statement.

Many AI agents we see on Twitter are just word predictors dressed in cool avatars. However, the market valuations of these AI agents are skyrocketing. GOAT has reached a valuation of 1 billion dollars, while aixbt's valuation has also reached about 200 million dollars. Are these valuations reasonable?

No one can be sure, but ironically, I feel satisfied with these assets I hold.

Data access is key

I have always been very interested in the combination of AI and cryptocurrency. Recently, Vana caught my attention because it is trying to solve the 'Data Wall' problem. The issue is not the lack of data, but how to obtain high-quality data.

For example, would you share your trading strategies for low liquidity small-cap tokens in public? Would you release high-value information that usually requires payment for free? Would you openly share the most private details of your personal life?

Obviously not.

Unless your private data can be protected at a reasonable price, you would never easily share this 'private data' with anyone.

However, if we want AI to reach a level of intelligence close to that of humans, this data is the most critical element. After all, the core characteristics of humans are their thoughts, inner monologues, and most secretive reflections.

But even obtaining some 'semi-public' data poses significant challenges. For example, extracting useful data from videos first requires generating subtitles and accurately understanding the context of the videos so that AI can comprehend the content.

For example, many websites require users to log in to view content, such as Instagram and Facebook. This design is common in many social networks.

In summary, the main limitations facing current AI development include:

  • Unable to access private data

  • Unable to access data behind paywalls

  • Unable to access data from closed platforms

Vana offers a possible solution. They break through these limitations by aggregating specific datasets into a decentralized mechanism called DataDAOs while protecting privacy.

DataDAOs are decentralized markets for data, and they operate as follows:

  • Data contributors: Users can submit their data to DataDAOs and thereby gain governance rights and rewards.

  • Data validation: Data will be validated in the Satya network, which is a network composed of secure computing nodes that ensure the quality and integrity of the data.

  • Data consumers: Verified datasets can be used by consumers for AI training or other applications.

  • Incentive mechanism: DataDAOs encourage users to contribute high-quality data and manage the use and training process of the data through a transparent mechanism.

If you want to learn more, you can click here to read more.

I hope that one day aixbt can break free from its 'stupid' status. Perhaps we can create a dedicated DataDAO for aixbt. Although I am not an expert in AI, I firmly believe that the next major breakthrough in AI development will rely on the quality of the data used to train models.

Only AI agents trained with high-quality data can truly realize their potential. I look forward to this moment and hope it is not too far away.