AI systems can now also beat humans in terms of creativity and can be turned into comprehensive productivity tools. However, this also poses challenges.
It's been more than eight years since I had the pleasure of holding the keynote speech at the 25th anniversary of Georg von Holtzbrinck School of Journalism. The title of my address was “Organic Journalism.” In that speech I said, “There's no time for stagnation; automated journalism is coming.” I could rest on my laurels now, and pat myself on the back for having correctly predicted a future trend. But that’s not the complete picture, and one also has to be able to admit mistakes. Because at the time I also said, “Our greatest unique selling point is organic journalism: Journalism made by human hands and human minds.” Judging by the developments we've seen in the field of generative artificial intelligence (“generative AI”) in the past 18 months, I have to admit: This may have been to bold a claim.
Automation is beginning to surpass us humans in the area of creativity – in scale, precision, and above all, speed. The major language models like GPT-3, LaMDA, Stable Diffusion, along with image and video generators like Dall-e, and code generators like GitHub's Copilot, are all creating results I wouldn't have believed possible just a few years ago. It's admittedly true that not everything they churn out is flawless. Just recently, tech company Meta had to take its “Galactica” language model off the market again after only three days. Galactica was producing fictitious research papers, but attributing them to actual, living scientists – spitting out abstracts full of false facts. Other systems, like Stable Diffusion, repeatedly struggle with racist and sexist content, precisely because the system learns from the billions of pieces of data it collects from the Internet.
And yet, what is happening in the field of AI development will fundamentally change the way we work with content. Just recently, research company OpenAI published a demo version of its ChatGPT conversation model. The system can write well-worded apologies and provide design drafts for home interiors, which can subsequently (with the help of Dall-e) be transformed into visuals. But the special feature of this system is its dialog mode, which makes it a sparring partner for business development.
ChatGPT writes poetry, academic and journalistic articles
For example, users can ask ChatGPT to develop an app for a given task. The system initially suggests the steps that should be taken. And if you then ask the AI to help write the code for the app in the Python programming language, it gets to work – and quickly produces the complete code. ChatGPT can also be used to write poems, academic articles, and journalistic reporting. The results are not always stunning. But they are only the beginning of what generative AI will eventually be capable of.
These AI systems can be turned into comprehensive productivity tools that can help us do everything everywhere all at once. Is this desirable? It will lead to us humans adapting more and more to the logic of machines. And machines don't care much about schedules, rest periods or overloading. It's different for us humans. Is it good for a society and economic system that is already operating at its limits to systematically allow those limits to crumble?
But generative AI does make dealing with many everyday tasks easier. When I ask ChatGPT to suggest an outline for a book about “Productivity in the Digital Age,” it gives me the results in less than 20 seconds. And even though the proposed book structure may not be the non plus ultra, it can still take me a step further. I can continue working with the AI's suggestions, adding my own thoughts and ideas – a step-by-step modification and refinement process that is at the core of nearly every creative process. With the help of language models, we are finally entering the age of human-machine emergence.
Generative AI will also challenge our view of humanity
But what are the implications of all this for the concepts of originality, authorship and copyright? There are a number of challenges society will have to contend with. GitHub's Copilot is currently the subject of a class-action lawsuit filed by a group of programmers because the system uses copyrighted code without attribution or licensing when generating its own new code. In such cases there are clear guidelines for human communication: One cites the original source. But as the numerous cases of revoked doctoral degrees over the past few years have shown, humans don't always reliably follow the rules, either.
Generative AI will challenge our view of humanity. The previous gods and goddesses of creativity (“humans”) will be dethroned. That hurts. On Twitter, a user asked Chat GPT what the philosopher Karl Popper would have thought about essayist and statistician Nassim Nicholas Taleb. The response generated by the AI is nuanced and worth reading. But even more worth reading is Taleb's own reaction on Twitter. He complains about the AI's “severe defects” and its “shallow thinking,” characterized by buzzwords and verbiage. It is a grave narcissistic slight to humanity that technology now might be capable of activities that previously were ours and ours alone. But relying exclusively on human, “biological” dominance might also prove fatal. It may turn into one of the black swans Taleb always warned about.