The year 2023 is not old yet, but it already has its first big hype. Social media is full of examples of clever or spooky chat histories using ChatGPT. The software seems to write texts independently and elegantly after being trained on a huge pile of snippets of language. There is also no shortage of announcements that the program and its competitors from Google and Meta would change the world.
But what’s the point in practice? How much genius does ChatGPT bring to everyday work? And how much work does artificial intelligence (AI) really save people? The SZ spoke to practitioners.
Building software with the Alfreds
ChatGPT from Open AI, Bing Chat from Microsoft, Bard from Google: Christian Ries from Heidelberg calls the chatbots enhanced with artificial intelligence “Alfred”. That’s the name of Bruce Wayne’s butler, aka Batman, and he’s always there to help the billionaire. Ries says on the phone: “I’m sitting here with three screens in front of me, on the left my digital dashboard for the project management dashboard, in the middle the development environment in which I program, and on the right ChatGPT and Bing Chat are always open.” The 28-year-old and three other IT experts build software for city administrations with their small company Jaimo, which is intended to help them with their social media appearances.
Chat GPT and Bing Chat are always open at Christian Ries.
(Photo: Christian Ries)
His “Alfreds” help Ries to quickly find the right building blocks for his programming code and to fix errors. That saves him a lot of time: “I used to not be able to program certain things. Now I can solve problems that I used to spend two days researching. Today it’s a prompt that I simply put to Alfred.” Prompts are questions or commands that users direct to the chatbots. It is important to know the limits of the program, says Ries: “It is clear that the AI talks shit quite often. But as long as it spits out code that I can fix quickly, I still saved a lot of time.”
Artificial intelligence is often described as a job killer. In Ries’ company, the Alfreds at least prevented a new hire, he says. “Last year we had great concerns at times that we could do everything. We thought that we would have to hire a fifth employee to deal with the backlog. But with the tool we are now much more productive, certainly by 20 percent.”
“He doesn’t have any emotional problems”
Mirko Kiefer reports from Bangkok. There the programmer develops apps with a small company. Kiefer says: “The bot replaces an employee in some respects. With precise instructions and goals, it works methodically, can do many things at the same time. And it doesn’t have any emotional problems either.” Kiefer is enthusiastic that the technology will make great leaps in the coming months: “What you need huge production teams for today, you can soon do in an afternoon as a one-man dig.” For example, produce videos seemingly out of nowhere: Some AIs can already do this because they have “seen” so many videos that they spin them together on commands from their users.
But what good does that do for Mirko Kiefer today? He says programmers have a big advantage because: “Such AI simply replaces a lot of tedious Google work. Writing code so far is 70 percent Google.” So even professionals looked for solutions when they got stuck – the information is all on the Internet, the only question is where. Programmers often have to go through “some marginal cases” on technical pages. “It’s all gone,” says Kiefer. “You can focus much more on the conceptual issues.”
He also uses the chatbot as a source of ideas when he sets up new websites to bring his apps to people: “For a new app that helps people with solar system installations, I only give the AI the goals and functions of the app as bullet points The AI decides which features need to be explained in more detail, which images should be included and where the potential customer can register.”
Inspiration but no emotion
So ChatGPT can help anyone who writes program code, but what about texts that are not intended for machines but for people? Giuseppe Rondinella works for a PR agency and says ironically: “I wanted to know: will I lose my job in a few years?” He now uses ChatGPT to create inspiration for press releases and posts for social media like Linkedin on a brief instruction from him. That saves time because writer’s blocks are eliminated: “The white sheet of paper loses a lot of its horror.”
But the 32-year-old also says: “I find the disadvantages much more exciting than the advantages. The texts lack personality – but a post in social media in particular has to live from personality. ChatGPT lacks any emotionality at all.” The same applies to originality. Because the artificial intelligence only learned from texts from the Internet that are already there. If a sentence reads brilliantly, it can only be coincidence, the results often tend to be mediocre. But that’s okay, says Rondinella: “I think it’s good if it doesn’t work well. Where would we be if the machines wrote better texts than the people?”