Imitation vs Innovation – will ChatGPT rewrite the future?

Imitation vs Innovation – will ChatGPT rewrite the future?

A couple of weeks ago Justin posted some thoughts about ChatGPT on LinkedIn:

“ChatGPT is “a grotesque mockery of what it is to be human”.

Not my words – that’s Nick Cave.

He’s got a sensational turn of phrase, as you’d expect – he also calls it “replication as travesty”. And “bullshit”.

What he really objects to is the idea that ChatGPT can write songs, or poetry. He sees those things as uniquely human – “songwriting…requires my humanness”.

And I have sympathy with him. I’m not anti-AI in general, nor anti-ChatGPT in particular. It has applications for sure, and I’m pretty awe-struck by how incredibly clever it is.

 

Oddly enough I feel much the same about recruitment technology and recruitment AI. It’s often very clever. It can shortcut processes and be effective. But it can’t – and I feel pretty strongly that it shouldn’t – replace human interaction. It shouldn’t be making decisions about people’s lives or replacing interviews.

When we work with companies we’re usually helping them find people who use technology to resolve business problems or take opportunities. Technology can add extraordinary value. But it needs to be harnessed, and focussed on the right problem, and used well. And that’s a job for a person..”

 

I have been playing with 3.5, and I know that the next version will be smarter and have even more fascinating applications. The version I’ve been using is reassuringly fallible and human – which might be its superpower from the point of view of lulling you into the sense that you might be talking to a human – but it pulls you up short when you realise that you actually can’t really trust it to come back with actual facts.

 

Anyone who has met me will know that I like books. I read books like other people drink coffee. Actually drinking coffee whilst reading books is my favourite pastime. I also like analysing, categorising, discussing, whatever you want to call it – give me a language generating AI model, and what I want to know is how it can talk to me about literature.

 

The results were mixed: but it was actually in the entertaining and spectacular failure that I felt like I had struck gold.

 

We started off in the 19th Century. Good old-fashioned canon territory. Books that frankly we all feel like we have read (even if we haven’t read them personally). ChatGPT is rock solid on Austen, Dickens and Hardy. Even able to make a decent fist of 1000 word essay comparing Hardy and Austen as social activists and use quotations from their novels to support the argument. So far so good.

 

I happened to be reading a book by a Scottish writer at the time – a book called O Caledonia by Elspeth Barker. It is a cracking book, originally published in 1991, and recently republished with an introduction by Maggie O’Farrell explaining why it is one of her favourite unknown books. A quick google search will show you that it is definitely not canon – but it is not obscure either. So straightforward question to ChatGPT – is this book a good example of gothic literature? The answer was an excellent example of on one hand…on the other hand. A beautiful piece of straightforward fence sitting: but it entirely made up the plot and the names of the characters.

 

In fact it did this 3 times. When challenged on facts (such as main character’s name, the historical setting or the plot) ChatGPT apologised profusely and then made up something else. As far as I can work out the plots it devised and threw out at me are not actual published works of fiction. What it’s actually done is give me three potential plots for totally new books. It’s accidentally creative.

 

Now I know, because ChatGPT told me, that it is a generative language model, it can’t read books, it doesn’t have feelings about them, the elements that it was juggling to continue the conversation had no intrinsic value to it in the way they might have to someone who had actually read the book.

 

Throughout the course of the conversation, I realised that I have had this kind of interaction before (albeit without the obsequious apologies) when talking to anyone else about a book I had read which they hadn’t (but felt they ought to have done). We’ve all done it – tried to bluff out an experience that hasn’t actually happened – seen that film, heard that song, watched that TV show, read that article.

 

All that said no-one has ever bluffed so blatantly, made up facts with such confidence or apologised so completely for their brazenness in the process.

 

To my mind generative language technology is an awesome tool and it is in the moment that it mimics human interaction most remarkably that it loses its value as a tool.

 

So what can it do well? And are all our jobs in danger?

After all, you are probably thinking, we don’t have many people in the workforce paid to have conversations about 20th century Scottish literature.

 

So far I have seen examples of short scripts developed in ChatGPT, summaries of technical writing, content of all kinds written with quite simple instructions and a reasonably playable game that won’t give Wordle a run for its money but was devised and coded by ChatGPT.

 

That final example is called Sumplete and the creator (or maybe co-creator) has posted both the game and the steps that went into its creation online. It is worth a look. The creation process definitely needed human guidance and interaction. As a unique creation rather than a creation based on external facts that need checking it also feels like an immediate use for the technology which will not require extensive googling to verify information.

I read an outstanding quote from Tom Goodwin today which addresses this very neatly: “To best furnish our kids for an abundant happy future they need to double down on topics that are the most human, not the most computer like.”

Ultimately what sets humans apart is their creativity. Their imagination. Their ability to relate to one another and the world around them. And their capacity to create and use tools. As the era of Generative AI dawns, here is an extraordinary new tool. It can’t replace our humanity, but it can help us to use it to the full. Jobs will change because of this technology. People who can harness it well will excel. And the boundaries of what can be done in many workplaces will expand.

People needn’t worry about competing with Chat-GPT. They should learn to harness it. 

Was Justin right then in his post? I think so. The computers are not after our jobs. This technology is there to be harnessed and used to grow businesses, free people up from repetitive tasks. Will it involve reskilling? Possibly. Will it require fact checking? Probably. Will we continue to value human interaction, doing business with people, trusting the creativity that comes from a team working together to achieve a common goal? Certainly.