How AI is undoing reality and how you could help put it back together.
Stories are easy to tell. It’s how they get traction and are taken up by others that’s the issue.
I did a rehearsal of a presentation to the academic council of the Pan African Heritage Museum (PAHM) around how AI can be used for film making for communities or groups lacking documentation, or even denied it. PAHM is a global cultural body that’s building a unique museum in Ghana.
My presentation coincided with a story executives in Ghana were taking note of, which I included in my talk from London. British newspaper The Guardian reported that “Industrial Revolution iron method ‘was taken from Jamaica by a Briton”. You kind of need that to sink in for a minute.
British Financier Henry Cort, dubbed the “father of the iron trade”, observed and lifted ideas for transforming scrap iron into wrought iron from Black metallurgists (Jamaican and West Africans) in Jamaica. “It was one of the most important innovations in the making of the modern world”, says Dr Jenny Bulstrode in her paper. The full story here.
Stories revealing minorities and under represented groups’ achievements ( Black and working class) throughout history are either ignored, brazenly hidden or plaintively lied about.
In the 1700s with race and class supremacy being coddled in Western societies, stories like the aforementioned would never see the light of day. Years later in the 1800s that’s how you get penny newspapers (costing a penny) in the 1800s, like the Coloured newspapers.
So, how could AI help retell that story factually, when the media doesn’t exist. One solution lies in using research papers to help dramatise events as realistically as possible, much as painters did before photography took over.
Trust was placed in the artist and events truthfully based on records. Hence for instance this image may be a worthy representation, based too on the idea that in the 1700s sketch drawers fulfilled the practice of today’s photographers.
But there was also a worrying component. Imagine it’s 2030. By then centaurs will be at work in many industries. Combination of humans and tech (See Kasparov advance chess 1988).
ChatGPT may be on version 12. It and Bard will be taken as gospel what it says. Humans supply the strategy i.e. direction of enquiry whilst the AI provides answers much like Smart speakers e.g. Alexa, which was based around Star Trek’s computer.
Chat GPT works by having access to an array of written data. Terrabytes worth. It treats words as part of a mathematical matrix in what users might approve within its ongoing training. Academic content (research) plays a significant content part in helping its language model achieve its output.
For instance if I say it’s raining cats and…? Bard/ Chat’s predictive step is aided by previous data input in which the likely word is “dogs”. If I say who is Henry Cort? Chat/ Bard work through data that already exist and reaches for strings of what’s likely been said of him. I used the word strings since Bard doesn’t comprehend words in the same way you and do.
The question then is what if that data doesn’t exist, what then?
UK Radio
This week I submitted an academic paper to one of the leading media journals. It charts the beginnings of BBC radio aimed at newly arrived first generation Black Britons to the UK in the 60s to the period of 1993. Specifically it links a radio show “Black Londoners” by Alex Pascall to Black London which I co-presented.
There are some interesting findings. For instance Black Londoners was a hugely popular show listened to by Black and White audiences and was warmly described by a BBC News Editor. It was the precursor to Black London which was aimed at second generation Brits ( 18–40 yrs) on the BBC’s London station.
When I asked Bard who presented Black Londoners. It rightly tells med Alex Pascall, but distressingly adds he passed away in 2017. Pascall hasn’t and is very much alive. When Bard was asked about Black London, Bard says The Ranking Miss P. presented Black London. Clearly it’s wrong and leaving my ego aside its symptomatic of wider issue.
And that comes back to academia and fact-based writings that are referenced, built upon, and eventually find their way into Chat GPT’s system. Built upon Dr Jenny Bulstrode’s paper about Jamaican metallurgists requires amplifying to a point that its presence impacts Bard’s output.
And that work is required now. If you work in academia or in journalism’s there’s a “reality” in process that requires addressing. I’ve been immersed in this field for quite a while, going back to the late 90s when I first became interested in code and platform building online.
Rehearsal done for PAHM, now it’s time to do something because by 2030 it may well be too late. My father’s story below using AI
