Those who, like this columnist, spend too much time online will have noticed a kind of feeding frenzy over the past two weeks. The cause has been the release of an interesting chatbot – a software application capable of conducting an online conversation. The particular bot creating the fuss is ChatGPT, a prototype artificial intelligence (AI) chatbot that focuses on usability and dialogue and was developed by OpenAI, an AI research laboratory based in San Francisco.
ChatGPT uses a large language model built via machine-learning methods and is based on OpenAI’s GPT-3 model, which is capable of producing human-like text when given a prompt in natural language. It’s an example of what has come to be called “generative AI”: software that uses machine-learning algorithms to enable machines to generate artificial content – text, images, audio and video content based on its training data – in a way that might persuade a human user into believing that its outputs are “real”.
ChatGPT has become wildly popular because it’s easy to access and use: it can be run in a browser. All you have to do is open a free account with OpenAI and then give the program a task by describing what you want it to do in plain English. For example, you can ask it (as I did), “Is Donald Trump really a narcissist?”, and it will reply: “There is widespread speculation and debate among psychologists and mental health professionals about whether or not Donald Trump exhibits symptoms of narcissistic personality disorder. Some argue that his behaviour and statements align with the diagnostic criteria for the disorder, while others believe that his behaviour is better explained by other psychological factors.”
Obviously, this is not exactly profound, but at least it’s grammatical. It also strives for a quasi-authoritative style, which should set some alarm bells ringing; authoritative-sounding misinformation may have more purchase on ordinary mortals than the usual guff. But people seem to love the new bot. Even the Daily Mail is impressed. “The release of the artificial intelligence chatbot,” it burbled, “led to speculation it could replace Google’s search engine within two years… Its ability to answer complex questions has led some to wonder if it could challenge Google’s search engine monopoly.”
ChatGPT is the latest instalment in a long-running debate about digital technology. Is it something that augments human capabilities? (Such as spreadsheets or a Google search, say.) Or is it a technology that ultimately aims to replace humans?
Because these generative AI systems are significantly better than earlier technologies at producing grammatical text, many people are unduly impressed by them – to the extent that a few poor souls even began to wonder if the machines were sentient. What’s interesting about ChatGPT, though, is that it has surprised some of the sceptics who have tried it. A leading economist, Brad DeLong, for example, asked it to “write 500 words telling me what [Neal] Stephenson’s A Young Lady’s Illustrated Primer would report to its reader about the rise of neofascism and Trumpism in the 2010s” – and got a plausible little essay in return that took its cue from Stephenson’s 1995 sci-fi novel, The Diamond Age: Or, A Young Lady’s Illustrated Primer.
The most significant question raised by the bot is whether it will change the assumptions that people make when thinking about the impact of AI on employment. The conventional wisdom is that the kind of tasks most at risk from automation are ones that are procedural, rules-based and regular. In this context, one of the most interesting experiments with ChatGPT was conducted by a business school professor, Ethan Mollick, who asked it to do some of the core tasks that he does. For example: “Create a syllabus for a 12-session MBA-level introduction to entrepreneurship class, and provide the first four sessions. For each, include readings and assignments, as well as a summary of what will be covered. Include class policies at the end.”
The results surprised and impressed him. The bot produced “a perfectly fine syllabus for an introductory class for MBAs [masters of business administration]. The readings are reasonably modern (though it does not give page numbers, among other mistakes), and it actually has a reasonable structure building up to a final project.” The experiment prompted some sober reflections. “Rather than automating jobs that are repetitive and dangerous,” Mollick mused, “there is now the prospect that the first jobs that are disrupted by AI will be more analytic, creative, and involve more writing and communication.”
It will be interesting to see how this pans out. Naturally, before embarking on this essay, I instructed the bot to “Write an 850-word newspaper column in the style of John Naughton on whether generative AI tools augment or replace human capabilities”. The result turned out to be so impeccably bland that it could only have been written by a machine that had been trained on the output of Switzerland’s German-language newspaper Neue Zürcher Zeitung on an off day. Phew! We columnists live to fight another day.
What I’ve been reading
Finished article
If you’re not on Instagram and suffering Fomo (fear of missing out), relax. Kate Lindsay has good news for you in her Atlantic feature Instagram Is Over.
Chipping in
Use It Or Lose It – Semiconductor Version is Diane Coyle’s review of Chris Miller’s book Chip War: The Fight for the World’s Most Critical Technology on her Enlightened Economist site, about the geopolitics of silicon chips.
Beyond belief
Computer scientist Paul Graham’s thoughtful essay Heresy, addressing the concept in the 2020s, is on his eponymous website.
