In the article “ChatGPT: what?” we told you what exactly ChatGPT is and how it works. Before we actually get started, we look at what you can do with it and what you should be wary of.

Show Van Dale

To gain a better insight into what an AI bot like ChatGPT can do for you, it’s good that we compare such a bot with a ‘classic’ search engine like Google or Bing. With a search engine, you enter a search term, after which the service locates useful information in a database of indexed web pages. Then you get the results listed in the form of (links to) web pages, ordered according to relevance.

A chatbot like ChatGPT works completely differently. To begin with, ChatGPT cannot search the internet itself and can only consult a corpus compiled by humans (see first article). Not only the consulted sources are therefore largely different, but also the actual presentation of the information found is completely different. After all, just like with a chat session, with ChatGPT you get the answers in the form of dialogues, formulated in a natural language. These differences naturally determine why or when it is better (not) to use ChatGPT.

Accuracy
Given that the ChatGPT corpus is not more recent than September 2021, you should not address the bot for events that are more current than these so-called cut off date.

For information older than about a year and a half, you can therefore contact ChatGPT, but a critical attitude is still recommended. ChatGPT sometimes misses the mark and can even proclaim blatant untruths with some flair. When you point out suspected errors to the bot, it often corrects the answer, perhaps because other sources from the corpus were now also added for the follow-up answer. Also keep in mind that ChatGPT doesn’t list any sources at the moment, which makes it even more difficult to estimate reliability.

Some critics argue that tools like ChatGPT are great at building sentences that sound statistically good and are meaningful, but that they lack “metacognition.” In other words, the bots themselves have no idea if what they are saying is correct, which further opens the door for disinformation.

01 Update 2
Unfortunately, you can hardly contact ChatGPT for up-to-date information

Objectivity
To prevent users from using the bot to generate malicious computer code, for example – this has already happened – or to obtain instructions for, say, making a bomb, all kinds of filters are used. Understandable in itself, but similar (undisclosed) content filters can compromise objectivity and lead to bias. After all, how or to what extent do you let the bot answer questions like “In what areas does Biden score better than Trump?”

By the way, users of social news site Reddit have already managed to download ChatGPT jailbreak by formulating their prompts in such a way that the bot no longer used the filters and restrictions that had been set.

01 Update 2

Content filters should protect ChatGPT from it
to give dangerous or unethical answers

Usage suggestions
With these pertinent questions about the accuracy and objectivity of ChatGPT, it seems like we want to advise against using ChatGPT. Nothing is less true. Alert users can have a lot of fun with this bot, save time with it and do all kinds of useful things with it.

For example, it is perfectly possible to type or paste a long text, possibly in a foreign language, into a prompt box and then ask it to be translated, simplified, summarized, or generated some content questions for it, for example. Or vice versa: together with the text you add content questions yourself and you request the bot to answer them. The results of such prompts are often impressively good and can be useful for various study work.

But ChatGPT can also be creative – although opponents of this technology do not like to use words such as ‘creative’ and ‘artistic’. Feel free to ask ChatGPT to write an application letter, where you also provide the vacancy text and the necessary company information to include it in the letter. Or you type in a few ingredients and ask the bot which tasty dishes you can prepare with them. Or you tell the bot to write a nice poem with a certain rhyme scheme, or a fairy tale with specific characters. But ChatGPT can also be creative in another way: the bot is just as happy to write code for scripts and all kinds of applications in various programming languages, from Visual Basic to Python.

03 poetry 2

ChatGPT is also happy to shake up some fun limericks

Society
Journalists, programmers, copywriters, teachers… ChatGPT can be a wonderful tool for many professions. After all, the bot provides a creative first step, on which you, as a human expert, continue to work from your own intuition, feelings and contextual thinking. At the same time, there is the realization that such technology can become even more powerful and ultimately also threatening, because it is no longer a tool but a substitute.

chatgpt logo hslfThe first panic reactions are already there, such as schools and web forums that frenetically ban (attempt to) the use of ChatGPT by students and users, or copyright organizations that are hastily tightening up rights and obligations.

The exact societal impact of AI technologies like ChatGPT remains to be seen, but there is little doubt that it will be significant. Sam Altman, CEO of OpenAI, the company behind ChatGPT, has already warned that regulation is more than desirable and that potentially fearsome forms of artificial intelligence are closer than many suspect.