AI Investing for (Investor) Dummies
Most often, I am asked for the list of tickers that benefit from AI rather than explain the technology, its use cases or revenue business models based on AI. I thought today might be a good day to give context to the AI boom, how it might play out and the yard markers to keep an eye on for potential busts. Are you ready? Watch this video first.
Large language models are essentially taking a massive amount of data, for example, a dictionary, a set of legal journals, medical records, the Internet or anything that is sitting in a format that can be captured, and then “learning” the material in order to provide “users” the ability to form useful queries. As we saw recently with the Google Gemini faux pas, the quality of the data and removing perceived bias is critical in claiming your model as “useful” commercially. A corporation that was planning to use Gemini as its LLM for Generative AI suddenly questions whether the data is “clean” and might generate mistakes or create backlash by end customers seeing the data from its generative AI results. Perceived woke, right or left leaning responses to questions kills a model’s credibility. Another example, a model that includes 2+2=5, the mathematical falsehood, in its training, needs to remove this from the data set to ensure it doesn’t mistakenly provide this information unbeknownst to users. Similarly, computer coding datasets used to build a “coding LLM” need to clean the data and train it properly, not a small task as mistakes are difficult to spot in lines of code. This does NOT mean the closed source (proprietary) Chat GPT has no such mistakes. Nor does it mean the open source (anyone can use) Llama 2 by Meta is clean for all use cases. But, the quality of the LLM built will be in the eyes of the paying customer which brings us to today’s discussion: AI Investing for Dummies.