The Philological Society of Great Britain is for those involved in the scholarly study of language and languages. In 1957, linguist J. R. Firth wrote an article published by the society entitled “A Synopsis of Linguistic Theory”. The article included a quote which has been repeated many times: “You shall know a word by the company it keeps!” Today, that concept has been modelled in computer systems using word vectors.
Large Language Models (LLM’s) use vectors with hundreds, or thousands, of dimensions for each token (part of a word) they analyse as part of predicting the words they output. Each vector is a long list of numbers which encode the statistical relationships between tokens (how ‘close’ they are to one another) to show the company they typically keep. It’s been found that they can be processed mathematically to find vectors with related word meanings, so that you can begin with the vector for “biggest”, subtract the vector for “big” from it, add the one for “small” and the result takes you to the word “smallest”. The link below give an example of a word vector.
Link: WebVector for “Rabbit”
The mathematical concept of thousands of dimensions expressing how close one thing is to another may seem rather strange to those of us just used to working within the four dimensions of time and space (length, height and width). But in our bonds with other people, we are familiar with a greater complexity. If we had internal ‘vectors’ to encode our relationships with others then they would have multi-dimensional depth. The vector might begin with how attracted we were to another’s outward looks. But then expand as our knowledge of them grew. Some of the traits stored in it would be clear to us: “I love her sense of humour”. Others we would find harder to pin down. And, of course, some could be there through weird or unfair bias: “I’ve never liked that name”. But whatever their source, each ‘value’ in our ‘vector’ for another would fine tune our sense of their distance from us – how relationally close we are.
However, the interesting question for the Christian is to ask how such relational vectors would need to be processed. LLM’s use their word vectors to find related words so they can generate output which seems readable and reasonable to their users. In similar fashion, humans are naturally drawn to being with and helping those we feel a bond with. But God tells us to refuse to act simply on the basis of how close we feel to another. Jesus commanded his people to love their enemies (Matthew 5:44) and linked that to the Old Testament law to love our neighbours (Leviticus 19:18) by telling a story about a Samaritan who comes to the aid of a Jewish mugging victim that fellow Jews have ignored (Luke 10:30-37). For Christians, our ‘relationship vectors’ aren’t the final guide to how we treat others.
And something wonderful happens when we learn to process this alternative way. We change. With the help of God’s Spirit, this algorithm of love makes us people who grow in goodness (Galatians 5:22-23).
All posts tagged under technology notebook
Introduction to this series of posts
Cover photo by Denley Photography on Unsplash
Scripture quotations are from the ESV® Bible (The Holy Bible, English Standard Version®), © 2001 by Crossway, a publishing ministry of Good News Publishers. Used by permission. All rights reserved. The ESV text may not be quoted in any publication made available to the public by a Creative Commons license. The ESV may not be translated in whole or in part into any other language.