I recently took something called the PXT Assessment. It’s one of those personality-style assessments often used in HR-Land. In years gone by, I would have been thoroughly sceptical and avoided taking such an assessment like the plague.
However, over the years, I have been persuaded to take a few different assessments (like Kolbe, for example) and have been pleasantly surprised by their accuracy. While there may have been some aspects here and there that felt a bit ‘off’, the results more or less ‘got’ me.
In January this year, I had the pleasure of meeting Alissa De Witt, who runs an amazing business called Executive Impact. Through reading Alissa’s book, The Coach Approach to Leadership I learned a simple but important fact that had previously passed me by. There are two types of assessment: ipsative and normative.
Ipsative assessments measure general orientations and preferences. While a normative assessment does this too, it also measures quantifiable traits which can be compared to a normed group. This means you get to understand not only general personality traits but also the level of intensity of each specific trait being measured.
All the assessments I had taken previously were of the ipsative variety. The PXT assessment, is normative, the type I had never taken before. And my goodness, what a difference. It absolutely nailed me in a way that none of the other assessments ever had.
Alissa uses this as one of the core elements of her approach to teaching leadership skills – and I totally see why. (Shameless plug: if you are in need of some help with your leadership skills and/or managing your core team, Alissa’s your gal!). And, by the way, in case you are wondering, it was because of working with Alissa on a custom GPT project, that she very generously arranged for me to take the PXT.
Anyway, why I am telling you all this in a post that’s supposed to be about building a software app? Well, two main reasons. First, the PXT assessment highlighted and underlined very strongly my intense disinterest in things mathematical (to put it mildly); but highlighted my strong interest in combing the creative, the technical and the entrepreneurial, which pretty much sums up the combined skill set required to do what I’m doing currently.
Now, a strong disinclination towards the mathematical and a strong interest in the technical might seem like a contradiction in terms. But I don’t think it has to be. And given that it is the mathematical that entirely underpins the existence of computing and tech, it’s possible to be in awe of the mathematical, without needing to deeply understand it or use it.
And I was reminded of this during one of my recent ‘mental model building’ sessions with ChatGPT, as I dug deep into the ‘mechanics’ of how a GPT and a Large Language Model (LLM) ‘talk’ to each other.
Given that math underpins everything in the world of computing, it should come as no surprise that it’s math that drives everything that happens when you interact with a GPT and its underlying LLM.
What blew my mind was the way that an LLM understands and responds to what we are saying to it. Essentially, the LLM holds an extraordinarily complex semantic ‘map’ which it uses to help it match, understand and respond to anything that we say to it. The ‘map’ is multi-dimensional and is so complex it’s impossible to convey it visually. However, like any map it is full of ‘co-ordinates’.
So, your input (i.e., what you say to or ask of your favourite GPT) gets converted into a set of ‘co-ordinates’ that can be used to find the appropriate ‘place’ in that multi-dimensional map that matches the meaning of your input. And remember this is happening multiple times over, at lightning speed!
Now, I’d run a mile from wanting to understand the calculation that converts our words into a ‘co-ordinate’ and then from ‘co-ordinate’ back into words. And by the way, just to add to the complexity, there’s an interim process that happens before that, where the words are converted into something called tokens.
But even with my disinclination towards understanding the actual calculations that are involved, I’m in awe of the process.
Something that seems so normal and every day to us (ask a question; get an answer) is achieved through an abstract concept that is simple and elegant; yet in its implementation is so complex it’s impossible to create an accurate visual representation of it.
Even for a person with a strong disinclination towards the mathematical, that’s truly amazing.
Until next time…
