My coworker enthusiastically shoved his phone in front of me and said, “You have to try it.
After I pranced around a warehouse in a unique backpack and frame that gave me super strength to move boxes, we were on our way home. Although a new future had just begun, it had felt unsettlingly futuristic.
OpenAI’s ChatGPT chatbot was operational in November 2022, and anyone could test it as long as it wasn’t already at capacity.
The follow-ups ChatGPT can handle, like “make it less formal” and “make me sound northern,” are astounding.
The company has a line of AIs it refers to as GPTs, which stands for Generative Pre-Trained Transformer, the most recent of which is ChatGPT.
Based on statistical probability, it functions like a supercharged predictive text, putting one word in front of another. Despite having no real expertise, it nevertheless manages to seem terrifyingly human and credible.
But if you push yourself further, play harder, and try to make it “take your job,” you might just land with a (comforting) bang.
A few weeks later, Radio 4 contacted to ask me to film a documentary about ChatGPT since it might even be helpful. This was at an Ethics in AI event at the University of Oxford (yup, that is what I do in my spare time).
The calls came in from several media outlets as the production simmered away in the background.
It seemed as though the era of “it’s too techie/no one understands it/people are afraid of AI/our viewers aren’t interested in technology” had abruptly ended.
The ability to test this sophisticated generative AI had altered the situation.
Everyone was in a frenzy as people experimented with it, academic institutions debated what uses they should permit, and professionals wondered what it meant for them. As is well known, it frequently hallucinated and got things incorrect, but the problem was that it also got things right.
I had the opportunity of listening to the world’s top academic analysis participate in the discussion of what AI means for us. This subject concerns to me and ought to matter to all of us, so I didn’t want to keep it to myself. And now, the discussion is widely shared.
When something is technologically feasible, we cannot simply let it happen. The place AI will have in our society, how we will interact with it and one another, how we will address its prejudice, and how we will direct its judgements must all be discussed, considered, and decided.
Sensitive Perception of Sensibility
The documentary’s production offered a changing target. As soon as we had bagged the visitors and addressed the problems, other situations arose. Some were hypothetical, while others caused problems.
We spoke with specialists, people who were embracing it artistically, people who were leery, and some people who were just curious about what it might mean for their careers.
We experimented with writing an introduction, questions, and a conclusion for the programme. You can interpret its work anyway you like, but what seemed impressive at first frequently deteriorated into the ludicrous.
“Good day tech enthusiasts…” is not my regular greeting of choice. Perhaps it should be, or perhaps AI is smart enough to know better, and perhaps we are already losing our capacity for sentience.
We have only seen this technology in its initial form. Indeed, it is stunning, and this is a significant turning point for AI.
I surely don’t discount the boundless opportunities and potential that will result. There will also be traps and several difficulties that we had never even considered.
This is a time of change, and the systems will only get better whether Dall-E is creating art, Google’s Bard, or any upcoming competitors.
Surprises and Errors
I’ve used ChatGPT-powered search as a Bing AI beta tester. “Knowledge is power” has never felt more accurate.
You can actually (virtually) converse when you combine technology that can mimic human speech with factual knowledge.
It enabled me to schedule some touring and organise a meal for 17 people. Although its link selection was unexpected and its advice was not absolute, it appeared sensible and helpful.
Finding a vegan restaurant close to a shooting site the following week may have been as frustrating for it as it was for me. How very human of me to think I was being overly picky.
Testing it came with the warning that we are learning together, and there may be some blunders and surprises as it has been known to reflect the spirit of its interrogator, but it did come with that proviso.
The subtleties of human behaviour and the complexity of human intellect cannot be accurately replicated by ChatGPT or any of its competitors. They are not humans, and they have never lived. However, they have a lot of options.
Notwithstanding my misgivings about it, ChatGPT was occasionally useful in producing the documentary.
It could not, however, distinguish between a casual remark and a “this ought to be in,” so I was unable to bounce ideas off of it or give them some thought. Thank God for Fiona, my incredibly human producer.
You can judge how well it performed the role of virtual producer. Yet, I don’t believe I will hang up my BBC lanyard just yet.