AI is currently finding its way more and more into various software. There is ChatGPT, which sometimes feels like an all-knowing human, DeepL uses artificial intelligence not only for its translator, but also for its new tool that improves written text, or Bunny.net provides an API to generate images “on the edge”.

But also Kagi, a paid search engine, has now published an experimental tool that can be used to summarize any article from the Internet. And I’m amazed at how well it works. Here is the summary of my last monthly review:

January was a busy month for the author, who split their time between two households. To stay fit, they opted for bodyweight exercises with Freeletics, and they also made progress on their AWS certification. They also found time to work on their blog software, implementing a plugin that displays the currently playing music via last.fm in the blog header. Despite the busy schedule, the author was able to give their girlfriend the time and space to study for her university exams.

And surprisingly, it can also summarize German articles in English. But this behavior doesn’t seem to be consistent, for example this post was summarized in German and the summary wasn’t as good as the other ones.

I bet Kagi uses the OpenAI API in the background.

For texts GPT-3 seems to be quite good, but I also used it for generating some code for me, with not so satisfying results. Simple programs it can do well, but for that you don’t need an AI to do it for you. It probably shows that AI in this context is just a “simulation of intelligence” (because it had such a large information base to learn from), as Nico Brünjes wrote in his (German) blog post.

Let’s see when Microsoft releases its new version of Bing.