#2 RESONATE: Risk, Simplicity, GPT Equity Research, Simulate the World, Discovery & Innovation
What I read, saw, watched, listened to, and thought about last week.
It’s All Happening!
AI continues to move F A S T! Some really incredible advances and info this week. I also ran across some good posts on risk and probability.
📺 WATCH:
Dario Amodei: Anthropic CEO on Claude, AGI & the Future of AI & Humanity | Lex Fridman Podcast
Yeah, it’s long, but packed with incredibly informative learnings.
How to Think About Risk
Oaktree co-chairman Howard Marks explores the true meaning of risk. He discusses the nature of risk, the relationship between risk and return, and misconceptions about risk.
A couple gems from this video:
We don’t know what’s going to happen even if we know the probabilities. We only have one outcome.
Even though many things can happen, only one will.
Sam Altman on what’s coming in AI: Get ready…
Chat bots
Reasoners
Agents that take on longer term tasks with multiple interactions with the environment
Innovators: Agent that explore a not well understood phenomena
Then, do all of these at the scale of a whole organization
https://x.com/slow_developer/status/1854934911230394573/video/1
📓READ:
4 probability concepts you need to understand as a trader to improve.
1. The power of large sample size.
2. Percentages don't add together.
3. Fat-Tails
4. Expected Value.
https://x.com/GoshawkTrades/status/1854264702546629005
Jobs on Simplicity
Can GPT Perform An Equity Research Analyst’s Job?
The research paper “Financial Statement Analysis with Large Language Models” by Alex G. Kim, Maximilian Muhn, and Valeri V. Nikolaev explores the potential of large language models (LLMs) like GPT-4 to analyze financial statements and predict the direction of future earnings, comparing its performance to human analysts and specialized machine learning models.
“…GPT and human analysts are complementary, rather than substitutes. Specifically, language models have a larger advantage over human analysts when analysts are expected to exhibit bias and disagreement, suggesting that AI models can assist humans better when they are under-performing. Humans, on the other hand add value when additional context, not available to the model is likely to be important.”
Key findings include:
1. Superior Predictive Performance: LLMs, especially when given “chain-of-thought” prompts to emulate human reasoning, achieved higher prediction accuracy (60%) than human analysts and performed comparably to specialized artificial neural networks (ANNs).
2. Reliability in Difficult Situations: The LLM outperformed when human analysts were likely to struggle, such as with small or loss-making companies.
3. Incremental Value Over Traditional Models: LLM predictions often added value to human forecasts and matched or exceeded the performance of logistic regression and ANN models, showcasing their flexibility in interpreting complex financial data.
4. Enhanced Financial Ratios and Narratives: LLM-generated narratives, focusing on ratios like operating margin and efficiency, contained useful insights that mirrored a professional’s analysis and informed its predictions effectively.
5. Complementary to Analysts: The study suggests LLMs could support financial decision-making and may shift the role of human analysts by acting as a high-efficiency tool.
Overall, the study argues that LLMs, with their ability to draw on vast amounts of general knowledge and reasoning, could play a central role in financial analysis and decision-making, complementing or even replacing traditional human analysis in some areas.
Tiny Troupe: Simulate People in the World.
Literally last week I was thinking about using multi-agents to construct a synthetic focus group where I could ask the group questions and have them reply as individuals with their own personas. Today (11/11) MSFT dropped this.
An LLM-powered multi-agent persona simulation for imagination enhancement and business insights. Can’t wait to play with this and see what I can do.
AI Leads to greater Scientific Discovery and Innovation
“Results suggest an intermediate view. In materials science, I show that AI can meaningfully accelerate invention. However, the model must be complemented by domain experts who can evaluate and refine its predictions.”
🧠 THINKING ABOUT:
Using sound frequencies for healing.
Agent synthetic focus group: I was thinking about this last week and then MSFT drops this (11/11). See above!
If you enjoy these posts, please share with friends and colleagues. And subscribe so you don’t miss out on what’s resonating!