AI in IT – use but don’t trust unreservedly. Report by Symetria

Read more
AI in IT – use but don’t trust unreservedly. Report by Symetria

As we reported in a recent post AI a growth opportunity for the IT industry or a threat? No Fluff Jobs report, the use of artificial intelligence (AI)-based solutions at work is a widespread phenomenon. ChatGPT is used by the vast majority of IT professionals in Poland. How do they use this tool? Is the information it gives them reliable? Symetria decided to survey this.

Shortly about the survey’s methodology

The survey “Symmetry between human and technology” was conducted in April this year on a group of 108 respondents – 60% of whom were representatives of the IT industry. Among the respondents were people with different professional experience, both executives and students, and the largest group were specialists. It is also worth noting that they were from the group of so-called innovators or early adopters (that is, being ahead of the mass users with their actions).

Almost every respondent (90%) had used ChataGPT for at least a month, and 29% communicated with it daily. Regular use of ChatGPT is evidenced by the fact that basically 1 in 3 respondents launch the tool via a shortcut in their browser, and more than 40% type the tool’s name directly into the address bar.

What is ChatGPT?

Before we move on to the results of the survey, a few words about ChatGPT itself. It is a tool based on artificial intelligence, created by the company OpenAI (one of its founders was Elon Musk). It was introduced to the public in November 2022. ChatGPT works on the principle of a chatbot – users type in questions and receive answers. Using it thus is similar to a dialogue and different from typing questions into a web browser.

ChatGPT vs web browsers

The greatest benefit of using ChatGPT comes from deepening your questions, that is, having a kind of discussion with it. According to the survey, the majority of respondents are aware of this – 22% do it always, and 44% often. The answer “rarely” was indicated by only 10% of respondents. Arguably, the ability to dig deeper into questions also influences the ability to get a satisfactory answer faster, compared to search engines. This is stated by 60% of the respondents, while three times fewer take the opposite view.

Almost all respondents (95%) type more complex questions in ChatGPT than in a traditional web browser, which should come as no surprise given the advancement of the tool. Respondents are aware of this – nearly three-quarters of them said that OpenAI’s solution understands the questioner’s intentions better than online search engines.

However, web browsers can still feel safe. There are a number of subject areas where respondents make greater use of them. ChatGPT wins in the areas of work tasks and hobbies, as well as knowledge of history and past events.

Check the answers!

60% of respondents are digging deeper into the knowledge given to them by ChatGPT on relevant websites. This action is also to verify the information given to them by the tool developed by OpenAI.

However, this percentage should have been higher, as as many as 86% had trouble finding the sources on which ChatGPT generated the answer to their question. Which should be a concern, especially for more complex questions, or those that are meant to help during professional duties.

The question about the source of the information is asked by less than half of the respondents (41%). And, according to Symetria’s survey, providing false information by ChatGPT is quite a frequent occurrence. It was encountered by as many as 77% of respondents!

Indeed, it can happen that if ChatGPT does not know the right answer it will give us the one it thinks is most likely, albeit wrong. These are known as hallucinations. More than 30% of respondents are unaware of this.

Knowledge of ChatGPT needs to be expanded

ChatGPT, in the answers given to the user, relies on existing patterns and is not capable of creating new, abstract ideas and concepts. Nevertheless, only 28% of respondents strongly disagree with the statement that ChatGPT can “think” abstractly. A quarter agree with such a statement (5% strongly agree and 20% tend to agree), and 22% have no opinion on this.

OpenAI’s tool is ad-free (at least for the moment), so it’s noteworthy that nearly 40% of respondents aren’t sure if ChatGPT contains hidden commercials, and 9% agree with such a statement.

Most respondents wrongly assume that they can refer to all their previous questions at any time – in fact, ChatGPT has a limited memory. It is also interesting to note that more than a quarter of respondents agree with the statement that if they lie to ChatGPT during a conversation, this misinformation will be remembered by it – in fact, this is highly unlikely.

The above data confirms that despite the increasingly widespread use of ChatGPT by Poles, they still do not have full knowledge of the tool. All the more so if we realize that the survey included innovators and early adopters, not mass users. ChatGPT can help a lot in daily duties, however, it should be used carefully and with full awareness of its limitations.

Do you need a reliable IT services provider?

Then, you are in the right place. We would be happy to talk to you about your next project.