Tuesday, November 21, 2023

A ‘hack’ for the human context window

 















‘Can it help me with due diligence?’  That was the question I heard recently as I was connecting with a friend on the topic of emerging AI.  It’s a topic I’ve been following for some time, but only recently would such a question start to seem viable. 

 

The reason we are here is due to computers being able to bring so much data together, in a manner that so much mimics neural architecture, that its responses to our inquiries can start to seem human.  My friend wanted to push that further – in venture capital, when you’re looking to invest in a company, the pathway towards the decision is called ‘due diligence’ – the process of developing a view on the opportunity based on relevant context.  That includes many topics – company leadership, financials, the industry and market, competitive advantages, and well beyond.  But what if we can just put those data points in a pot and POOF – out comes good investment decisions?

 

At this point, that concept is science fiction – but there is an interesting pathway ahead, connected to what is called the ‘context window’ in AI terms.  The context window is the amount of information the computer can take in at once prior as part of any response.  Think of the difference between a preschool age child and an adult, in terms of the amount of context that can be held at one time.  In the case of the computer, this context window has been increasing rapidly.  Last year it maxed out at the rough equivalent of 5 pages of text.  As of last week that number is about 300 pages – enough to fit an entire book in, ask the computer for an in-depth summary, and get it immediately – in fact, possible to ‘interview the book.’ 

 

Humans primary system of context is through the logical brain, which can hold much less context (less than a page), but we augment this by connecting to lots of data – books, conversations, numbers, etc.  We take it in with the attempt to integrate, within our limited context window, in the most efficient manner possible.  And that system is pretty solid.  In terms of making investment decisions, it is still more capable than the 300 pages for AI to work with.  However, we can expect to see the context window number for computers to increase exponentially in coming years – moving from 1 book to 10 books to 100 books number of context, and beyond.  It’s important to name that this level of context wouldn’t be as an external ‘reference’ (which is already possible to near infinite levels) – our equivalent of consulting an encyclopedia – but rather as integrated, equivalent to referencing our own life story.  It seems our context window supremacy is going to be short-lived.

 

There is another system of context in humans – the nervous system.  It’s been articulated and studied in a variety of contexts and domains, but seldom in a manner that places it in any way level with logic – generally having it seen as untrustworthy at best and malicious at worst. 

 

But increasingly, that view is changing.  Folks like Antonio Damasio in neuroscience, Peter Levine/Gabor Mate/Staci Haines in psychology, and lots more.  There is a shift afoot – we are not thinking machines; we are feelers who think. And it makes sense – the feeling of what happens carries orders of magnitude more information than the thoughts.  Here we have a context window that can be much, much larger than what a computer can take in.  This can be leveraged, but is also fraught.

 

There’s a reason many of us have been told in our lives that we can’t trust our emotions when making decisions – that logic and rationality are the gold standard towards sound decision-making.  There are many contexts where our nervous systems can be hi-jacked!  Additionally, we have built an entire culture disconnecting and dulling ourselves with regards to our bodies, feelings, and sensations.  The pathway of re-membering, from what I’ve seen, is non-linear – but really benefits from ethical grounding, a general sense of other-orientation, and an orientation towards self-inquiry/agency.  With those ‘preliminaries’, here is a teacher of the Kagyu Tibetan lineage describing the process of ‘befriending our beautiful monsters.’ 

 

There is a world in psychology called Somatic Experiencing, which is an articulation of how one can connect with feelings and sensations as a pathway to even bypass thinking and narratives entirely, on a path of psychological healing.  The reason for this is multi-fold, but it includes the understanding that leveraging that larger context window can be an upstream way to shift personal narrative, without even necessarily engaging the narrative.

 

For the last several years, I have taken these insights, combined with practices connected to eastern Dharma, towards leadership and service, with friends in the spaces I work in.  Specifically, towards relationships, conflict, decision-making – places where the limited context window of the thinking mind becomes more evident.  And I’ve found clarity emerge, again and again.  Self-driven, based on curiosity towards discovery, and over time, a greater sense of fluidity with the embodied self.  Instead of decisions being made, pathways forward emerge.  Rather than conflict being something to be avoided or beside the mission, it shifts towards being an opportunity for a dance of self-discovery.  The body has more space.

 

And along with these types of high leverage uses of the nervous system, comes due diligence!  I’ve also seen, again and again, more easeful capital allocation decisions when engaged with in a somatically informed manner.  Leaders who feel aligned with their path forward, such that the second guessing simply dissolves – regardless of what comes later.   It is possibly inevitable for AI to grow its context window massively in coming years, but the tremendous context possible to be derived from the feeling of what happens remains a meaningful offering germane to those humans among us who cultivate the sense.