Come what come may,
Macbeth, Shakespeare
Time and the hour runs through the roughest day
Earlier this year (March, 2023), when I was preparing a lecture about ChatGPT, I asked it the fundamental question of life:
Who am I ?
The mystics would say that we must look beyond the roles and labels we have in life – mother, wife, friend, computer scientist, etc., – which lead to the social identities that give us a sense of belonging and purpose, and look to our innate essence and pure existence to find the answer there.
ChatGPT said:
I wish I could answer that! Unfortunately, I don’t have any information about you. But I’d love to get to know you better! What are some things you’d like to share about yourself?
I found the exclamation marks disturbing and my disappointment more so even though I know it’s just software and has no ‘intelligence’ at all, so I asked Eliza the same question, to see how far we have come AI wise.
Eliza which was one of the first natural language processing systems, or chatbots, created in the 1960s by MIT professor Joseph Weizenbaum who named his software after Eliza Dolittle in George Bernard Shaw’s Pygmalion.
I used two online simulations: Playclassic games Eliza and Fullerton’s Eliza. The responses to my: Who am I? were:
Why do you ask?
And:
What is it that you really want to know?
These answers seemed more appropriate somehow and yet, I know that ChatGPT is a lot more sophisticated, technically speaking. It has constructed the answer word by word, sometimes syllable by syllable. Eliza’s responses are just full sentences pulled from memory printed onto the screen – rather like pulling scraps of paper with answers already written on them, out of a hat – but they resonate because the answers match the tone of the question. ChatGPT’s response doesn’t. There is nothing wrong with it. It is just not what I ‘expected’ given the nature of the question.
We all have expectations and desires in any interaction because we are human, and when we are just seeing words on a screen, regardless of who/what wrote them, we bring so much more to the conversation than is actually there. We bring ourselves to that gap between us and them and so it is easy to anthropomorphise code and imbue it with an intelligence that it does not possess. Research shows as humans we give our loyalty to brands and algorithms in exactly the same way that we do to the humans in our lives based on how much we trust them.
So when a bit of software responds to my philosophical question with two exclamation marks and a breezy answer, it feels like it has shut me down and denied me space, rather like when I say something vulnerable, funny, or clever, and people just pass right over it because they either didn’t get it, didn’t want to get into it, or just wanted to get the interaction over with me as soon as possible. It’s hard not to feel rejected and since we never have a raw experience, in that our new experiences are coloured by our old ones, I respond in the same way to software as I do people. ChatGPT works by prediction based on probability. Eliza works by pattern matching and stock questions generated randomly and yet, here I am judging them as if they are human, because connection and transference give our lives meaning.
After that, I asked more prosaically: Who is Ruth Stalker-Firth?
I wanted to see what ChatGPT had to say in order to see how it constructed its information and where had it got its information from?
First though, I asked Eliza, who said:
Does that question interest you?
ChatGPT the shredder
On this webpage my About page tells the world what I think it needs to know. ChatGPT has read every page and post on my website everyday, sometimes several times a day. It steps through all of them each time (as I write this it’s been round three times this morning) which I first noticed in 2021 during lockdown when I wondering what to do about those brute force hackers who try for hours at a time to crack the login over and over. And so it was that I saw bots from Unix machines in Oregon owned and run by Microsoft, step through every post, every word, on every hour, and there were so many of them, they were impossible to block. I had begun feeding the machine of ChatGPT. ChatGPT on a daily basis takes all of my original ideas, consumes them and spews them back out, randomly, to tell a new story, something that we used to celebrate, after all, the action of recombining existing things is the definition of originality, academically speaking.
Naively, I thought that ChatGPT, when asked who is Ruth Stalker-Firth, would use the information from my page. Of course, after delving into how it works, I know that it cannot just copy and paste, like a human, as it is not programmed that way. Instead, it shreds and de-contextualises everything it reads and then rewrites something else completely different word by word, predicting by probability what word comes next, without ‘understanding it’s meaning’ so even when asked the same question, it will have a brand new answer, as it doesn’t remember anything.
Sometimes I am a physicist, or a digital strategist, more often than not I am an influential thought leader who has given talks at TEDx Women in Tech and Agile Manchester, and written a couple of books with cool titles, I googled them and it turned out that they were real titles and had been written by someone else. The connection: we have both had cancer, though ChatGPT didn’t mention that, it didn’t ‘remember’ because it can’t.
In the same way, ChatGPT generates artwork based on someone else’s without credit, but, OpenAI will not take responsibility for that intellectual property theft because it would cost them dearly, not least of all by forcing ChatGPT to be coded to cut and paste such details, we would then see that it doesn’t really work at all so instead they want us to believe it has run riot and they can’t control it. As I write this I see Sam Altman has been ousted from OpenAI seems like he is the fall guy for ChatGPT’s plagiarism chapter. (Update: Three days on and he has been reinstated.)
With all that said, looking at what it had written about me was like a form of fortune telling, someone predicting an alternative, yet possible, reality, in which I give talks and write books, was dazzling and inspiring in the way it is when we see ourselves through someone else’s eyes, until it reminded me of Macbeth.
Revisionist History
Machine learning and Shakespeare has been on my mind for a while as in the Westworld blogs, I talk about their AI machine called Rehoboam, which predicted people’s lives and created the exact environment in which they had to fulfil them. This is not a new trope, it was in the James Bond film Tomorrow Never Dies, where the villainous news tycoon Jonathan Pryce writes headlines and then gets his henchmen to fulfil their prophesy. In computing we have Moore’s Law , how fast transistors get made. It has been, for over 50 years, both a prediction and a business target.
Serendipitously, everyone in our house is studying Macbeth for GCSE English, so this summer, we went off to the Globe Theatre to see it live. It was brilliant as we were standing in the ‘cheap seats’, the courtyard, right by the stage so when the actors came on and off they passed by us, and created an immersive experience. However it was jarring experience, because the Globe Theatre is, by design, meant to look old and the changes they had made to this production were to bring it up-to-date so my experience of the play was confusing.
The three weird sisters were three men in hazmat suits wearing masks. I’m all for balancing up the genders but was at a loss as to why Duncan, now a woman, wore a dress and a crown whereas Macbeth wore a bulletproof vest. The Thanes were females in trouser suits, bringing news to Duncan, as if they air hostesses. Everyone lived in castles but were in modern day clothes sourced from M&S and Boden, and some of the text was changed and serious scenes were played for laughs. It reminded me of when we ‘improve’ someones’s way of working, we introduce ‘progress’ with a computer. Information is simplified, summarised and then forgotten.
To replace the weird sisters with three fellas in the forest loses the reference to the Three Fates who weave the destiny of humanity, either intentionally or not. They are female archetypes and the opposite of that which is man-made (or created by a scheming murderer), something Shakespeare tries to play on later when only someone not born of woman could kill Macbeth. Now I have had two caesareans, and I can tell you my body knew both times it had given birth and honestly, they weren’t untimely ripped, I was labouring forever. When you remove the Fates weaving destiny you remove the question: Was Macbeth predestined to be king and would it have happened anyway? Is that not the tragedy? Blinded by ambition Macbeth went on a killing spree to become the king that he would have become anyway because he couldn’t wait.
Standing as I was in the thick of it for the first time, it struck me how quickly he met them and how determined he was to make their prophesy come true even though it was just three randomers on the heath. I’m always meeting randomers giving me unsolicited advice. Why didn’t Macbeth just say thank you and be on his way?
The story goes that Shakespeare wrote the witches in because James I believed he had been cursed by witches and went on to write Daemonologie which promoted witch hunting. When your fates become fellas the story becomes the age old one of patriarchy: Kill the women and children and grab the power for yourself.
Apparently, #irl, Macbeth was a good and just king, but us humans never let the truth get in the way of a good story rather like ChatGPT shredding and stealing nowadays it’s no longer wrong or invalid, it is hallucinating, rather like when Macbeth is haunted by the ghost of Banquo and Lady Macbeth says he’s unwell.
Perhaps ChatGPT is unwell and rather like the courtiers who suspect that Macbeth is really not doing anything good, they say nothing and go along with that fiction, because their livelihoods and lives depend on it.
