Family offices generate a massive amount of gravity. Investment opportunities, innovations, trends and professional networks orbit a family office like stars and planets swirling around black holes at the galaxy’s center.
The families who thrive and survive are the ones who learn how to harness that power for the good of the family, philanthropic causes, innovation sectors and the investors that fund them, as well as for people and communities. The ones who don’t learn how to harness that immense gravity are the ones who eventually turn into a black hole, having consumed all resources, familial relationships and purpose.
You cannot invest today without considering the zenith known as AI — which has the tech and investing world buzzing with possibilities, many of which seem significant. One parlor trick, though, is to ask someone raving about it to explain what AI is and what it truly technically offers, and often we receive one of two responses: “machines that think like humans” or just immense silence.
When you break it down, three things are commonly referred to as AI:
- Machine learning (ML): While this is in a similar field and the baby brother of AI, this is not what technologists consider AI. ML is a computer ingesting 10,000 images of a hot dog to see whether it can recognize a hot dog. Machine learning has immense strength and value, but be aware of entities selling ML as fully formed AI.
- Massively trained large language models (LLMS): Massively trained LLMs are where the big boys of Silicon Valley are currently playing. We see models trained around 7 million to 9 billion parameters — yes, with a "B." This means that they have the appearance of being highly functional.
The problem with this structure is that quantity is not always quality. Look at Gemini from Google. While remarkable, it also told users that part of a complete diet was “eating rocks.” Or to use nontoxic glue on pizza to make the cheese stick better. Entities like ChatGPT, while an exceptionally powerful LLM, have many issues with where they source their data and still some problems with accuracy and permissions. Open AI Whisper, which has transcribed millions of medical conversations, has been found to “hallucinate” in around 1% of conversations, especially during long periods of silences. These models will continue to be a significant part of AI in the future but also a big part of our discussion on ethics, liability and bubbles of investing. - Correctly trained LLMs: What do I mean by correct? When you break down the training of data, especially that of an AI model, you get semantics. Not how we conversationally use the word but what is also called semantic web, ontologies, taxonomies and knowledge graphs. Unbelievably, ontologies or the mapping of information and knowledge are usually attributed to around 2,300 years ago, when philosophers like Aristotle tried to start labeling life and everything around us. Well, guess what? That methodology came in handy when starting to create really specific databases. Knowledge mapping and ontologies are the secret sauce to making AI extremely good and making a lot of money. Do you know how Amazon makes its product recommendations with unerring accuracy based on your purchase history? Knowledge graphs and ontologies. Do you know how to create an amazing AI? Create an amazing knowledge graph and ontology first.
So, what does this mean for family offices? Well, there are two essential things:
- The better the knowledge graphs and ontologies, the better the AI; the better it helps cancer treatment outcomes, investing strategies, and business growth; and thus the more revenue and social good it can do. So do not be fooled by machine learning wearing a mask as AI. Pop the hood and check out the engine or have a mechanic (technologist) you trust to do the same.
- As many researchers have noted, many models have limits, data bias or developer bias. As AI becomes more widespread and adopted by mainstream companies and consumers, the statistical risk of derailing the long-term goals of a family office grows, not only from implementing possibly faulty AI solutions but also from competitors. If single- or multi-family offices are not yet working on an AI operations (AIOps) plan, they are behind institutional investors and corporations.
So, what happens when two black holes collide? They entwine and can become a supermassive black hole. Whether that jump-starts a new rich galaxy or a massive spot devoid of all light depends on the right conditions. Family offices have some conditions needed to create or assist with this AI development. Just a few categorical sectors of family offices that would benefit from a proper AIOps department, service or plan are:
- Portfolio growth and planning assistance
- Asset tracking and management
- Inheritance strategy
- Generational training and assistance
- Smart health care and longevity assistance
- Philanthropic strategy and management
This leads us to the question: How can my family office create the right conditions? The answer is adaptive — no, seriously, adaptivity. Personal informatics is the study of all data categories about a human being. It is about how we can map, store and analyze every scrap of data on a human being to help improve quality of life. The same applies to family offices. Think about it like family office informatics (FOI), a topic for a future article.