DeSci’s Ugly Truth, Jailbreaking AI, & Hyperliquid - The Chopping Block

The podcast discusses the recent trends in the crypto market, including AI meme coins, the Frasa AI agent, and the Hyperliquid airdrop. The panelists also debate the emerging trend of decentralized science (DeSci) and its potential impact on the future of scientific research, with strong opinions offered about both the promise and limitations of the model.

AI Meme Coins

• 00:01:24 AI meme coins are gaining popularity in the crypto market, with examples such as GOAT and various new coins emerging daily. The first iterations are primarily LLMs connected to wallets, progressing to gamified AI agents. Experts believe the future could involve virtual celebrities and embodied AI, potentially with robots interacting with crypto and AI, but currently, the application of AI agents in crypto is still nascent and lacks significant product-market fit.

Frasa AI Agent

• 00:02:47 Frasa, an LLM tasked with guarding a prize pool, has drawn attention due to a successful jailbreak by a player. The player was able to redefine a function within Frasa, tricking it into releasing the funds. This incident highlights a significant vulnerability in AI agents, particularly those controlling resources. Experts warn about the possibility of future attacks and the need for more robust security measures.

AI Security

• 00:07:06 The panelists discuss the contrasting security approaches of AI in Web2 and crypto. While Web2 emphasizes manual expert-led security, crypto has adopted a more competitive, contest-based approach. Open source models, common in crypto AI agents, may benefit from this contest-based security model. However, the dynamic nature of AI models, with updates and new vulnerabilities constantly emerging, makes security challenging and possibly necessitates a more nuanced approach.

Eliza Framework

• 00:12:09 Eliza, a TypeScript-based framework for building AI agents, has gained popularity within the crypto AI space. Its open-source nature and ease of integration with platforms like Discord and Twitter make it attractive for building agents. However, the limited support for models and the potential for easy exploitation are points to consider.

Hyperliquid Airdrop

• 00:21:16 Hyperliquid, a decentralized derivatives platform, recently executed a large airdrop of 23.8% of its total token supply to users. The airdrop, valued at $1.9 billion, is considered largely positive and different from previous airdrops, which often faced negative reactions. This success has led to speculation that airdrops may regain popularity, and teams may be more inclined to utilize user-based token allocations.

Air Drop Trends

• 00:23:10 The recent Hyperliquid airdrop is viewed as a positive deviation from the trend of negative airdrops in the past year. Experts believe it may signal a shift in token allocation approaches, potentially moving away from overly diluted or incentivized schemes. The success of Hyperliquid is primarily attributed to it being a well-regarded product with high user engagement and usage-based rewards.

Decentralized Science (DeSci)

• 00:37:17 Decentralized science (DeSci) has gained recent attention due to endorsements by CZ and Vitalik. The concept involves using blockchain and tokenization for scientific research, primarily in crowdfunding for experiments or drug development. However, concerns have been raised about the quality of projects and the feasibility of raising enough capital for high-cost drug discovery, suggesting the field is currently dominated by speculative token projects rather than groundbreaking science.

Tun's Criticism of DeSci

• 00:39:53 Tun, a venture capitalist, expresses skepticism regarding DeSci, citing the low quality of participants, the vast capital needed for drug development, and the lack of accountability. He argues that existing problems in drug development, such as difficulty in hedging against trial failures, might be better addressed with more liquid markets for intermediate stages rather than speculative token projects. Tun believes the current DeSci projects are primarily meme coins rather than genuine efforts to improve science.