A group of researchers from the University of Oxford has developed a new tool focused on training Artificial Intelligence and Machine Learning systems in business skills.
Called JAX-LOB, it is the first GPU-accelerated limit order book (LOB) simulator of its kind.
The researchers explained that traditionally, LOB simulators have used CPUs, but due to the importance and central role of these order books in the financial system, it was necessary to develop a new system that would take advantage of modern computing hardware and parallelism to improve efficiency and facilitate the execution of multiple tasks at once.
According to the researchers, the new tool enables AI training in trading skills 7 times faster than traditional CPU-based LOB simulators.
They also commented that the new tool allows for increased “capacity to accurately and efficiently model LOB dynamics,” which is extremely valuable in the financial market. As an example, the researchers explained that this could allow a financial firm to offer better services, or allow a government to predict the impact of financial regulation on the stability of its overall system.
A system for commercial optimization
JAX-LOB focuses on optimizing trading tasks through AI to add a layer of quality to financial markets.
As Oxford researchers explained in the study “JAX-LOB: A GPU-Accelerated limit order book simulator to unlock large scale reinforcement learning for trading”, its innovative training system enables the training of AI trading agents, through the use of Reinforcement Learning (RL), to perform different automated trading tasks, such as market making and trade execution or trading.
In addition to RL, JAX-LOB also makes use of the machine learning agnostic framework JAX, developed by Google, for high-performance machine learning research.
In the paper, the researchers commented that GPU-based LOB systems are a largely unexplored topic, so the JAX-LOB design will also help create and open new avenues of research that will allow for further application of RL to trading operations.
AI will be able to conduct its own financial experiments with JAX-LOB
Jack Clark, co-founder of Anthropic and co-chair of the AI Index at Stanford University, called JAX-LOB a supertool designed for super-intelligent AIs. He noted that this LOB simulator can “process thousands of books in parallel with significantly reduced processing time per message.”
On the importance and potential of this new tool, Clark highlighted that JAX-LOB seems to be the exact type of tools that artificial intelligences of the future will be able to use to conduct their own financial experiments. “We can expect that the primary users of these tools will be the AI systems themselves!” he commented in his latest blog post. newsletter on AI research.
What is Limit Order Book or LOB?
Limit order books or LOBs are electronic records of orders or requests to buy and sell an asset that require the trader to specify a price and quantity.
These order books are widely used in traditional financial markets and also in the world of cryptocurrencies.
Cryptocurrency exchanges use this type of order book to coordinate market activity by matching liquidity providers with traders of an asset.
As mentioned above, LOBs require traders to specify both the order they want to execute, whether buy or sell, as well as the price at which they want to buy or sell the asset and the quantity of the same. Once the order is placed, it is recorded and published on the LOB so that, when it matches a compatible order from another trader, it is executed, thus completing the transaction.
Continue reading: Cathie Wood: The convergence between Bitcoin and AI can transform the corporate sector
