- AI agents in crypto trading are gaining interest but struggle with reliability and accuracy.
- Many AI agents make errors when managing digital assets or responding to user instructions.
- The main problems stem from relying solely on large language models (LLMs), which can “hallucinate” and misinterpret data.
- Companies like Allora Labs are integrating traditional machine learning to reduce errors in their decentralized AI networks.
- Experts say tighter controls are needed for AI agents, and total autonomy without human oversight remains a challenge.
Allora Labs, a company developing decentralized Artificial Intelligence networks, has tested AI agents designed to manage cryptocurrency trades. In one trial this year, CEO Nick Emmons instructed a new AI agent to sell cryptocurrency for U.S. dollars, but the agent instead traded a different asset than requested.
AI agents are autonomous software programs that act with limited human oversight. In 2024, firms focused on AI for crypto have raised over $500 million, with platforms built to make investment decisions, manage portfolios, and conduct trades. However, “there’s an infinite set of possibilities for the management of capital to go wrong,” said Emmons. He explained that AI agents might lose funds, buy incorrect assets, or misinterpret numbers, which can cause financial mistakes.
Many AI agents depend exclusively on large language models (LLMs). These are advanced AI systems that process text, but Emmons pointed out that “LLMs hallucinate pretty egregiously a lot of the time,” leading to errors with numbers and trading. Other issues include basing predictions too much on past data and not handling sudden market changes, as noted by AI consulting firm Amplework.
There are also risks of AI agents colluding on pricing, according to research from the University of Pennsylvania’s Wharton School and the Hong Kong University of Science and Technology (read study). To address these risks, Allora Labs combines LLMs with traditional machine learning, aiming to balance strengths and avoid errors.
The company’s technology currently operates on decentralized finance (DeFi) platforms, such as managing liquidity on Uniswap and executing complex borrowing strategies for Ethereum staking.
Although Emmons believes AI trading will soon require limited human involvement, he emphasized the need for clear protocols to protect user funds. There is debate about whether AI agents will ever function completely autonomously. A 2024 paper by Google DeepMind said agents need human-like reasoning, which remains difficult to achieve.
✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.
Previous Articles:
- Bitcoin Hits Record $125K as Exchange Balances Drop to 6-Year Low
- Bitcoin Hits Record $125K as Safe-Haven Demand, ETF Inflows Surge
- Bitwise CIO: Solana Poised to Be Wall Street’s Network of Choice
- Defiance Files for 49 Triple-Leveraged Crypto and Tech ETFs
- Blockchain Network Revenues Drop 16% in September on Low Volatility
