Three questions and answers: How is it programmed with artificial intelligence?
The hype surrounding ChatGPT couldn’t be greater – actually always a sign that skepticism is appropriate. But when used correctly, the AI language models are already a real support in the job – even for IT professionals. The iX title authors Philipp Braunhart and Gerhard Heinzerling report on what is already possible and where the models still have problems.
Philipp Braunhart is a developer and consultant in the field of machine learning and co-founder of ingenio ai GmbH. He programs, designs and builds AI solutions for direct use in the fields of bio, agricultural and medical technology.
dr Gerhard Heinzerling did his doctorate in 1999 on the question of how words are stored in the brain. He then worked as a SAP consultant and is now employed at Arineo in the field of image recognition using AI.
What are the advantages of a dedicated programming AI like Copilot over general systems like ChatGPT?
Philipp Braunhart: In the background, both systems are related to each other via the pre-trained GPT models. However, Copilot is specially adapted to the needs of software development. It is itself packaged in a lot of additional software, so that it can be integrated directly and easily into the IDE, for example, and works there like an autocomplete function. ChatGPT can also code, but works more in dialogue, like a colleague you can ask for advice and discuss the solution with.
What should be considered when dealing with the AI if you want to get the best possible programming results?
Philipp Braunhart: Before you consult the AI, you should ask yourself the question: “What information would another human programmer need to do the task I’m asking the AI to do?” After all, copilot needs the same information. The easier the logic is recognizable and the more specific and consistent the naming of variables, functions and classes, the better it works. Meaningful comments also help.
Mr. Heinzerling, let’s take a look into the crystal ball: Despite all the hype surrounding ChatGPT, the tool is far from perfect. Will the newly announced GPT-4 magically solve all of our AI problems?
Gerhard Heinzerling: Absolutely not. GPT-4, like its predecessors, is a great tool. However, with limited options. An example: GPT-4 is able to write a short story or a poem. But we are miles away from a complex novel. The direction is right and there will certainly be great progress, but it has yet to be made.
Mr. Braunhart, Mr. Heinzerling – thank you very much for your answers! The title spread of the current iX, which is available now, provides even more information on the professional use of artificial intelligence in everyday working life.
In the “Three Questions and Answers” series, iX wants to get to the heart of today’s IT challenges – whether it’s the user’s point of view in front of the PC, the manager’s point of view or the everyday life of an administrator. Do you have suggestions from your daily practice or that of your users? Whose tips on which topic would you like to read in a nutshell? Then please write to us or leave a comment in the forum.