Artificial intelligence (AI) has changed the nature of engineering and software development as well as the search habits of the world. Fewer people are clicking on links to referenced articles and taking the AI summary for granted. Usually, it’s pretty good, but sometimes it’s just AI slop that also seems to be permeating the content on the internet.

Of course, this article is a blatant push for you to regularly read or subscribe to Electronic Design Weekly. However, it’s a discussion about using AI, too, especially chatbots that have been integrated into every search engine, browser, and most applications. The key is the human in the loop.

The Challenge of Using Chatbots for News

The challenge these days is handling the massive amounts of information being dumped on us. The internet played a part by making all of this information available, but AI-generated content is significantly increasing the amount. Unfortunately, the resulting content is often duplication and AI slop.

This isn’t to say that generative AI hasn’t been useful. I use it daily for searches, summaries, and programming. I get rather frustrated with some results, though, where I have a summary but not links to the source material unless I ask for it.

Do you use a chatbot to summarize your emails or articles?

Try it on an issue of Electronic Design Weekly (EDW). I did.

Scanning the results took just about the same time as scanning it, since EDW is really just the table of contents with links to articles written by real engineers — not AI-generated stuff. The big difference was being able to click to an article or video and see something other than another AI summary.

Do You Trust Your Chatbot?

A lot comes down to whether you trust your news sources and the chatbots that may be providing you with the summaries. Dealing with the overwhelming amount of content, both real and AI slop, is a challenging task even for editors. Bias and manipulation are watchwords when it comes to AI because no one knows all of the material used to train a model. It’s why AI tools are used.

Bias and manipulation of the training data will affect the outcome of any interaction. The question is what types of bias are included, whether intentional or unintentional. Keep in mind, the guardrails that many are trying to place on models have a similar effect.

So, what do you trust?