When new technologies impose themselves on societies long habituated to older technologies, anxieties of all kinds result.
Marshall McLuhan
Anybody who writes for a living, is, or should be intrigued by the recent development of ChatGPT, the AI chatbot developed by Musk’s research laboratory, OpenAI and launched in November last year. If you are at all curious, you would, like millions of others, have signed up to experience the remarkable ability of this tool to generate coherent scripts, essays, texts, creative writing and even code in seconds based on a few simple prompts.
First one is free, of course and for that reason, the system is often in a state of denial of service as eager fingers everywhere interface keyboards to prompt scarily rational, highly versatile, and (mostly) contextually accurate answers rolling out in staccato bursts.
It’s not sentient, but it might feel that way because apart from drawing on a very comprehensive (and soon to be even more comprehensive) set of large language models, it was further tuned by human trainers to refine its responses as well as to learn from feedback provided by its estimable user base.
Despite its limitations (apparently it isn’t immune to bias, being factually incorrect, spouting occasional nonsense and hallucinating), it’s a significant milestone on the journey for computers to complete communications tasks that were otherwise deemed the near exclusive domain of humans.
It’s certainly got the chattering class chattering louder based on the sheer amount of content on news platforms about it and in the short two months it has been available to the public, it has also inspired a lot of YouTube techbros to put up videos guiding users how to monetize its usage (another being making videos about it on YouTube, of course).
I’d certainly put it on the same footing as say, the arrival of desktop publishing and social media and we all know where those have taken us. But, in one sense it’s slightly more capable of misuse, simply because all those other phenomena are already so ubiquitous and woven into our lives. A lot of pundits have pointed to the potential for example, of plagiarism in academia while others see clear applications for criminal syndicates to exploit it by automating phishing and other confidence trickery owing to the highly credible texts it can generate.
On a societal level, also, many have voiced concern that a good many white collar professional jobs will be replaced by similar generative chatbots in future, and given the upward curve of all technology, I am inclined to believe them. How long that takes is another issue but humans are remarkably adaptable so for now, we should classify generative AI as a new tool for us to work and play with rather than a doomsday machine — for now at least.
As a phenomenon, it’s the first shot of one tech giant across the bows of another. Much has already been written about it being an existential threat to Google’s highly lucrative search engine business. It’s good to remember though that similar engines have been constructed in Google’s vaults and that Microsoft has suggested it might build ChatGPT into its proprietary search engine Bing — think old paperclip Clippy, the Office suite assistant on steroids.
From my own perspective, I fear two things arising from generative AI — first, a weakening of the coming generations’ ability to express themselves clearly and concisely in written form. This follows a pattern that started with the loss of cursive script to the use of block print, the keyboard and now smartphone and from reading traditional paper to audio books or no book at all (my 14-year-old declared outright that reading books for pleasure was of no interest as she looked up from her TikTok feed).
Automation in general has made us lazy and this is just one more advance move in that direction.
Second, the blurring of reality as we move deeper into the uncanny valley. Humans have always been fascinated by mimics in nature, from the archetypal mockingbird to the more directed abilities of parrots and corvids. Language is what makes humans, humans and when we see its analogue in other animals we are naturally intrigued. Our shared evolutionary path also gives us a clue; the gene required for the proper development of speech and language in humans FOXP2 is found in many other vertebrates, where it plays an important role in mimicry in birds (such as birdsong) and echolocation in bats.
Generative AI engines are essentially gifting computers with the power of mimicry in natural language and now written and transcribed language too. It’s questionable whether we need to create robots physically in our own image but integrating this extra layer of fluid intelligence on top is pushing the boundaries ever closer to creating simulacra. The idea has been explored endlessly in science fiction (Westworld and the film Ex Machina being noteworthy fictional explorations) but it feels like we are approaching a new paradigm in the real world, one that might be reached in decades.
It’s always tempting to paint devils on the ceiling but it’s also important to ask those ‘what if’ questions in good time. I’ll leave this piece with a ‘what if’ posed by Neil deGrasse Tyson on one proposed path for evolution. Make your own conclusions.
