Is anyone doing any work or research in leveraging chatGPT's capabilities using open source knowledge graphs? I've come across the use of semantic models via knowledge graphs and would be a beginner in the subject. My current use of Logseq seems like it can be useful as training data for chatGPT to experiment with various components of creating an algorithmic trading system. However my use of pages is rudimentary and not really using the properties function is a forward way that can be extensible. My use is more like an outliner and hasn't been structured as a true machine readable knowledge graph. Financial Industry Business Ontology (FIBO); modules- Market Data Ontology (MDO) and Securities Trading Ontology seem promising but I don't know if using these models are really necessary unless API's for trading platforms use them. Any references to where I can learn more about this subject would be appreciated. via ChatGPT; The STO provides a framework for representing different types of financial trades and transactions, including the concepts of price movements and chart patterns. Within the STO, the class sto:MarketDataPoint represents a data point in a time series of market data, which can include price and volume data for a given financial instrument. To represent the concept of a 5-minute bar's higher high or lower low, we can use the sto:MarketDataPoint class and add additional properties to represent the relevant data points. For example: sto:highPrice: A property that represents the highest price value within the 5-minute bar. sto:lowPrice: A property that represents the lowest price value within the 5-minute bar. storeviousMarketDataPoint: A property that links the current sto:MarketDataPoint to the previous data point in the time series, which can be used to compare the current high and low prices to the previous bar's high and low prices. By using these properties, we can represent the concept of a 5-minute bar's higher high or lower low within the FIBO model. This can be useful for various types of analysis and decision-making tasks in the financial industry, such as technical analysis and algorithmic trading.
Cool little tool for knowledge graphs, link to GitHub if you want to tinker https://instagraph.ai?ref_id=UXGEYTNUP
I have been thinking of taking time off trading beyond tbills/index and learn to build something like this with a vector database and word vectors. The immediate trading use isn't something I really see but I feel really stale on ideas. This and the biopython tutorial is on my list. It is a huge tutorial that is so deep into sequences. Vector database though seems like a tool I just need to get. chatGPT gives a good summary for Ontology in this context: So an ontology always has semantic meaning? Yes, the fundamental characteristic of an ontology in the context of information and computer science is that it provides semantic meaning. The primary goal of an ontology is to establish a shared and common understanding of a domain by formally representing concepts, relationships, constraints, and rules in a way that is both machine-readable and interpretable by humans.
https://www.elegantthemes.com/blog/business/how-to-make-an-ai Along the vein of getting data into an ai... there is a (data) format converter named Pandoc. They call it a swiss army knife. It is open source. https://pandoc.org/index.html https://github.com/jgm/pandoc
Pandoc, fuck yea. I would contend though, first, this idea of thinking in vectors..... Thinking in vectors is maybe on the idea of the enlightenment? Maybe not....maybe not not
I took a shallow look at Word Vectors a few years ago. I came to a conclusion it is a twist on good ol' bit-twiddling. That's not a negative, just a personal observation... simplistically correct, and misunderstood on the back end. Ai prompts are a monetizable thing!! A growing industry in itself. Prompts are all about word vectors. **IMO**, Ai is like beta/vhs, blockchain, EV charging plugs, diet/zero. For now, let MSFT, GOOG, AMZN, and whatever new entrants duke it out for format. It will take time if there is to be one all knowing VGER. But all of them need prompts!! https://youtube.com/shorts/n2N7SuPDDmw?si=2iqCcX_G2uNU3dXl
Over todays morning coffee I typed the phrase from the above YT link into Bings CoPilot Ai. Here's the response... Obviously, Co-pilot does not know about the mnemonic, or has not had its bits twiddled for branching to a (mnemonic) context vector as a response. Just as interesting, imo, a basic (Bing) search of the same phrase (the YT link) produces over 2.6 MILLION results. Have a nice weekend.
Slightly off-topic, just found out Logseq can also be like a computational notebook like Jupyter / Observable (One has to search on Youtube for it, links are disabled) I still can't thank you enough for the tip to get into Logseq. It's been such a game-changer for me and I constantly send good vibes to you because of it.