It’s the year of endless hype. The potential of the new technology appears huge, but warning voices are being ignored. We are, of course, talking about NFTs. NF-what? In 2021, this was supposedly the next big thing, listing digital media on a blockchain. Looking back, this seems ridiculous, but at the time the glorification was reminiscent of the current debate about AI.
But unlike back then, language models and neural networks are not a speculative bubble. The technology has substance, its benefits are real. Nevertheless, the comparison helps to classify the excitement that Chat-GPT caused a year ago. Market researchers at Gartner are currently paying attention to generative AI the peak of the so-called hype cycle. This means that with almost all technologies, excessive expectations are followed by a trough of disillusionment until the plateau of productivity is finally reached.
The disillusionment has begun
But generative AI is far from that. Only a fraction of the more than 100 million users who tried Chat-GPT in the early months have firmly integrated it into their professional lives. Most people who have played around with Google’s Bard, Microsoft’s Copilot, or other chatbots in the past year are still unsure how the technology can help them in their everyday lives or at work.
As research assistants, language models too often produce nonsense. When it comes to facts, you have to check the answers yourself – you can also ask the search engine of your choice. Current chatbots also do not provide original or creative texts. They imitate and reproduce the material with which they were trained. If you want the result to not sound like it was spat out by a text machine, you’ll have to make an effort yourself, for better or for worse.
A sparring partner
The best way to imagine the chatbot is as a kind of sparring partner. The first answer almost never meets expectations. Together we refine the result, make suggestions for improvement, ask for relaxed wording or a serious tone, ask for shortening and keep trying out new prompts. AI doesn’t do the thinking for you, but it can transform quiet pondering into a dialogic process. That helps some people.
Chatbots can also be useful when editing your own texts or summarizing someone else’s. If you are writing a professional email in a foreign language, are not entirely satisfied with an application text or want to condense a scientific paper into bullet points, you can ask the assistant for advice. The more precisely you formulate the instructions and the more context you provide, the easier it is to do something with the result.
“We tend to overestimate the impact of a technology in the short run and underestimate it in the long run,” futurologist Roy Amara said more than half a century ago. There is much to suggest that the same thing will happen with generative AI. Google and Microsoft are just beginning to incorporate the technology into their products. Nobody knows how people will work with AI support in five years.
AI help today
Although the productivity plateau is far from being reached, there are AI tools that can help today. Two of them come from Germany and have proven themselves for years. Long before Chat-GPT was released and sparked the current boom, its developers relied on AI to help people with tedious tasks: translating and revising.
DeepL makes Google look old when it comes to writing in a foreign language with relative ease or translating coherent texts into German. Against Language tool Microsoft Word’s spelling correction seems outdated. The writing assistant not only reliably finds typos and grammatical errors, but also improves language and style.
Both programs cost money with full functionality, but even the free versions are more practical than all current chatbots.