Brett Harrison, the former CEO of FTX.US, believes that artificial intelligence (AI) can make coding algorithmic trading strategies easier. His new venture, Architect, officially started in January 2023 with a $5 million seed round from Coinbase Ventures and Anthony Scaramucci, among others. Architect’s platform has been available for select clients in beta mode since May.
Harrison said that the hardest part about creating such algorithms is the initial step, going from zero to one, because it requires specialized skills and knowledge that not all trading firms have in-house. To address this, Architect is leveraging OpenAI’s application programming interfaces (APIs) and combining them with their internal trading APIs to make it possible to use natural language probes to generate trading strategy code. For example, using Architect’s platform, traders could set up software that buys bitcoin on Coinbase (COIN) when its price drops below what it is on other crypto exchanges.
Architect is also considering using large language models (LLMs) for advanced market indicators, or actually embedding LLMs directly into the trading process. LLMs such as OpenAI’s GPT-3 and Google’s BERT are artificial neural networks trained on large datasets of language, such that they can understand and generate human-like texts.
“We do think that AI is going to be a critical component of most businesses going forward, and we didn’t want to miss the opportunity to be able to leverage AI tools for our users,” Harrison said. He noted that the success of the trading strategies depends on the team behind it, not just the generative AI used to build it. “You definitely need to review the strategy, make sure it’s working, tweak it, test it,” he said.