Over 1.000 comments
Generative AI for Asset Managers Workshop Recording
Unleash the Potential of Generative AI in Asset Management: Discover, Learn, and Apply!
The recorded content from our enlightening 2-day workshop held from September 30 to October 1, 2023, is now available for purchase. Hosted by industry stalwarts Dr. Ernest Chan, Dr. Roger Hunter, Dr. Hamlet Medina, and featuring an insightful keynote by Dr. Lisa Huang, this workshop meticulously explores the deployment of Large Language Models (LLMs) in asset management, particularly focusing on crafting robust discretionary trading strategies.
Learning Goals
Generative AI for Asset Managers is a 2-day online workshop to demonstrate how we construct a discretionary trading strategy using a LLM. We will demonstrate how asset managers and traders can use Google’s BARD to turn unstructured data such as the audio feed of the Federal Reserve’s Chair’s speech into high frequency trading signals and backtest such strategies, all at minimal cost. Participants can explore and experiment with variations and improvements on the basic code, as well as other use cases of LLM for asset management.
Workshop Speakers
This workshop is hosted by Dr. Ernest Chan, Founder and CEO of Predictnow.ai, Dr. Roger Hunter, Chief Technology Officer at QTS Capital Management, and Dr. Hamlet Medina, Chief Data Scientist at Criteo. We are honoured to be joined by Dr. Lisa Huang, Head of AI Investment Management at Fidelity Investments who will present as a keynote speaker.
Host
Dr. Ernest Chan
Founder and CEO of Predictnow.ai
Host
Dr. Roger Hunter
CTO of QTS Capital Management
Host
Dr. Hamlet Medina
Chief Data Scientist at Criteo
Keynote Speaker
Dr. Lisa Huang
Head of AI Investment Management, Fidelity Investments.
Workshop Overview
Day 1:
Delve into Generative AI and LLMs in Asset Management
- Comprehensive exploration of Large Language Models (LLMs) like BARD, ChatGPT, and their manifold applications in the finance sector.
- Harnessing the power of LLMs to structure discretionary trading strategies efficiently.
- Hands-on session: Transforming unstructured data into actionable high-frequency trading signals.
Day 2:
Advanced Techniques and Real-World Applications
- Fireside Chat with Lisa Huang
- Extensive coverage on prompt engineering and strategies to mitigate risks inherent to LLMs.
- Augmenting trading strategies with nuanced sentiment analysis using LLMs.
- Hands-on exercise: Backtesting trading strategies and a brainstorming session on LLM’s potential to revolutionize asset management practices.
Workshop Outline
01
Large Language Models (LLMs) & Generative Pre-trained Transformers (GPT)
- Introduction to LLM: BARD, ChatGPT, and other large language models
- Typical Applications of LLMs
- How LLMs work
- Using BARD/PaLM on the web through their API
02
Building Applications
- Overview of Prompt Engineering
- Building applications such as text generation, summarization, etc.
- Few-shot learning with BARD
- Introduction to embeddings
- Overview of the BARD embeddings API and its usage
03
Risks Associated with LLMs
- Understanding main risks with LLMs, such as, hallucinations, bias, consent and security
- Methods for reducing the risks of Hallucinations, such as, retrieval augmentation, prompt engineering, and self-reflection
- Methods to detect and address hallucinations, including reinforcement learning from human feedback (RLHF) and model-based approaches
04
Using LLMs for trading Federal Reserve Chair’s speeches
- Why we chose the BARD family among the many available LLMs
- Evaluating BARD’s native performance
- Improving performance with embeddings
- Worked example: computing sentiment ratings on public companies using embeddings
- Test data: Video archives of the press conferences of the Federal Reserve Chair.
- Backtesting a discretionary trading strategy using the sentiment output of a LLM.
05
Deploying LLMs in Production
- Best Practices for Deploying LLMs in Production
- Overview of alternative generative models such as ChatGPT, BART, Cohere, Alpaca, etc.
