What I learnt from my conversations last week on the latest on AI in trading

1. More on the Chinese AI arms race – Following High-Flyer, a Chinese hedge fund that effectively integrated AI into its trading strategies, other Chinese fund managers are accelerating their AI adoption. Firms such as Baiont Quant (https://www.linkedin.com/company/baiont-quant/) , Wizard Quant (https://www.wizardquant.com/en), and Mingshi Investment Management (https://www.linkedin.com/company/mingshiim/ )are leveraging AI to process market data and generate trading signals, with local governments, particularly in Shenzhen, providing substantial financial support – https://www.reuters.com/technology/artificial-intelligence/after-deepseek-chinese-fund-managers-beat-high-flyers-path-ai-2025-0
2. Why Manus is a game changer – designed to function independently, Manus reduces the need for constant human oversight which makes it more suited for complex workflows like stock analysis, earnings call summaries, technical indicators, and portfolio stress-testing. By making the process more autonomous, this creates additional requirements such as the need for compliance checks but moves the integration of AI in trading, a step closer. Read more here on the differences between Manus and ChatGPT – https://www.perplexity.ai/search/please-provide-a-detailed-brea-subq2hLUR_Ca86nTgb1mbQ#0
3. Open AI is responding with the introduction of their tools to build AI Agents. (https://openai.com/index/new-tools-for-building-agents/?utm_source=theaipulse.beehiiv.com&utm_medium=newsletter&utm_campaign=openai-s-five-new-tools-to-build-ai-agents&_bhlid=98599fe4152cf8c4f9b61a473cd965376e1ac1a0). But greater use of agential AI brings new challenges, leading to increased interest in Neuro-symbolic AI. This is useful for agential AI because it combines deep learning (neural networks) with logical reasoning (like following rules and making clear decisions). This mix helps agential AI systems make better decisions, explain their actions, and adapt to new situations more easily. For example, a trading robot can use data patterns to predict stock prices (neural) while following risk rules to avoid bad trades (symbolic). This combination makes AI agents smarter, safer, and better at handling complex tasks with less data. Read more on neuro symbolic AI here (https://www.perplexity.ai/search/please-provide-a-detailed-brea-subq2hLUR_Ca86nTgb1mbQ)
4. The ability to model using less data avoids the growing risk associated with synthetic data sets. A recent paper from Charles le Halle highlights the challenges in using AI models to simulate financial data where accurate data generation is crucial for the success of modelling. The paper points out issues like relying on too little data or creating too much fake data can lead to poor results. It also explains that general AI models may miss important details needed for designing strong investment strategies. To fix these problems, the authors suggest a better method for generating realistic financial data and provide a way to spot weak models. Read more here – https://arxiv.org/abs/2501.03993
5. Along with data the other critical element for greater integration of Agential AI in trading will be API integration. By enabling different systems to communicate and share data in real-time will be the only way AI agents can talk to each other effectively ensuring real-time risk management and compliance. APIs will also help integrate new AI solutions with older systems, ensuring that trading operations run smoothly. Read more here – https://rapidaddition.com/what-is-api-integration-in-finance-and-why-do-you-need-it/

As usual please let me know what you liked, didn’t like and what you felt was missing

Many thanks

Rebecca