Human + AI > Human or AI
The financial trading community had thrown its proverbial hands in the air at the idea of being able to reliably predict movements in stock prices. Both manual approaches combining Financial Algorithms and Charts with Human Intelligence; and automated approaches using AI / ML (Artificial Intelligence and/or Machine Learning) or Financial Algorithms have failed to consistently deliver above average market results.
Historically, predicting stock price movements by modeling Business Fundamentals, Economic Environments and Investor Psychology was not possible using Charts (even Candlesticks) and relatively simple Financial Algos / Technical Indicators. AI alone, also lacked the business and economic long-term insight (or for that matter Natural Language Understanding) to make accurate predictions. Computing costs also presented barriers to entry.
However, Semantic Brain has demonstrated a revolutionary architecture that delivers accurate predictions; and is continuing to enhance its ability to help Investors better manage their trades and portfolios. We are able to achieve these breakthrough results by uniquely integrating Financial Algos, AI / ML, Charts and Human Intelligence. We are harnessing the power of Deep Learning, Parallel Processing and GPUs to substantially improve accuracy while reducing costs and increasing performance.
This weekly journal is authored with the intention of sharing our journey, our accomplishments and industry trends / innovations.
Semantic Brain: August 16 to 22, 2021
During the week of August 16 to 22, Semantic Brain demonstrated breakthrough results in predicting stock price movements . We launched the Future Top Movers functionality (which identifies stocks that are expected to move more than 5% within five(5) trading days / one(1) week) into production.
On August 16th, Future Top Movers functionality identified 17 equities as most likely to have at least 5% price movement within the following 5 days. These 17 equities were selected from a sample size of 2000 (i.e. RUSSELL 2000).
94% of the equities identified attained a closing price which moved more than 5%
100% of the equities identified attained an intraday high / low which moved more than 5%
Semantic Brain has clearly demonstrated that its AI has the ability to help Investors select the best stocks that deliver the highest yield. Beyond this we are also providing Investors the tools necessary to interact with the Data and AI such that Human + AI can achieve far superior results.
Key Accomplishments Summary
Our key accomplishments for the week of August 16 to 22, 2021 include:
Future Top Movers: This alert or scan enables traders to identify and time high yield opportunities (these are generally not a part of the watchlist or portfolio). 94% of the stocks identified on August 16th had a closing price that changed more than 5%, while 100% of those equities reached had an intraday high / low surpassing a 5% change.
Due Diligence: Semantic Brain product Asperios delivers 1000x Investment Analysis (See video). Users can select from Future Top Movers list or any equity (in many key exchanges) and then analyze earnings, dividend payments, financial ratios as well obtain hard to find information on insider transactions, short selling, and institutional / fund holder activity. Investors can subsequently add these equities to watchlists and portfolios.
Ongoing Monitoring & Management: Asperios is designed to monitor watchlists and portfolios in an ongoing basis (scan every 30 / 5 mins and issue real-time alerts). The application can then identify risks (accuracy rate of 75% to 85%) enabling users to proactively manage larger portfolios.
Additional Details
Future Top Movers for August 16, 17 & 18 are identified below
All Future Top Movers reach intraday highs / lows exceeding 5% change
Fulcrum Therapeutics delivered significant price change (up more than 5%)
Angion Biomedica delivered significant price change (down more than 5%)
Cortexyme Inc. delivered significant price change (down more than 5%)
All except for 1 equity realized gains / losses of more than 10%. Predicting magnitude and direction are important.
Semantic Brain: Future Efforts
Semantic Brain is embarking on a long-term journey to "Augment Financial Intelligence", and aspires to eventually augment multiple forms intelligence. We believe that the following near-term agile R&D pipeline will help our customers achieve superior results.
Directional & Momentum Predictions: Deliver accurate predictions on direction of stock price change using momentum and directional indicators. Semantic Brain is working on improving the accuracy of these models, and we are planning to release this into production in September.
Broader Scans: Scan more markets and equities to identify better and more investment opportunities
There are 53K global equities, our coverage is currently limited to 3000 US equities
Real-time Alerts: Alerts users on historical anomalies and deviations in real-time.
Intraday Predictions: Semantic Brain's predictions presently use Daily Data. We plan to collect data every 30 mins (and subsequently 5 mins) to make more frequent and accurate predictions.
Expand Financial Instrument Coverage: Predict price changes for Crypto, Commodities and Currency Pairs.
AI & Cyborg Journey
Semantic Brain relies on a number of technologies (e.g. Predictive Analytics, Explainable AI, Chatbots) and enabling platforms (e.g. Large Multi-screen Platforms, AR/VR in future). We tie these technologies and platforms together to create Human and AI feedback loops which result in Superior Intelligence that is greater than Human or AI.
This section is dedicated to highlighting some of the above industry trends and innovations. We plan to eventually use Semantic Brain's Frameworks and Solutions developed for investing / finance in many other areas (e.g. marketing, medicine). Sharing our thoughts on industry trends and innovations is also our way of supporting and encouraging the community.
Reward Is Enough
Google's Deep Mind recently published a paper that hypothesised that AGI (Artificial General Intelligence) can be achieved using Reinforcement Learning alone - https://deepmind.com/research/publications/2021/Reward-is-Enough . While we do not believe that a single Neural Network architecture can deliver on AGI, we do believe Reinforcement Learning
1) can be used to stitch together many neural network architectures to deliver AGI
2) Reinforcement Learning could be used to augment Human and AI feedback loop
An interesting opinion piece by Ajit Jaokar on "Reward Is Enough" can be found at -
AI Designed Chips
Google and other organizations (e.g. Intel, Qualcomm) have started using Reinforcement Learning to optimize Chip Floorplanning to optimize PPA (Power, Performance and Area). These companies have been able to achieve superior results within hours (Chip Floorplanning is a process that generally take weeks or months). This is a major breakthrough, and there is also the promise of these Chips improving the SOTA (State of the Art) in AI.
Humans are able to perform complex tasks by breaking the problem down into smaller manageable components (build blocks, modularity, strategy). By taking this approach humans do lose out on some information. AI on other hand optimizes the overall solution as a whole, we have seen this finance when predicting stock price movement.
Even with AI Designed Chips AI will not be replacing humans. Humans + AI will augment each other.
Tesla AI Day
It is clear that Elon Musk / Tesla are doubling down on AI. Tesla has taken on massive Engineering Efforts in
Neural Networks
Computing Hardware - D1 DOJO Chip, Inference at the edge
Data & Labeling: 3D Images (3D vectors) / 4D Video (4D vectors)
Tesla Bot
While many will question timeline of fully autonomous cars, and viability of Tesla; these investments undoubtedly will move AI & Robotics forward. Innovations in niche transportation applications, driver safety and household robotics are likely.
Recommended content includes this video by Lex Fridman
DISCLAIMER: This is not investment advice. This article was drafted for information and educational purposes only. Investors are advised to perform their own due-diligence or seek input from professional investment advisors.