Can artificial intelligence become sophisticated enough to take over our jobs?

At this point, it doesn't seem like we need to fear the darkest of predictions, but we do need to accept that things will change. Significantly. This was evident at the recent panel discussion AI: Myths and Truths, held at the Let'sGrow tech conference in Pula in mid-May. The conference focused on artificial intelligence (AI), featuring nearly 30 speakers.

Myths and truths about AI were discussed by Goran Mrvoš (Infosit founder and CEO), Davor Brenko (Istra Tech CEO), Siniša Miličić (Faculty of Informatics in Pula Dean of Teaching and Students), Dejan Rajković (CMG Digital Director of Technology), and Oskar Vujičić (Penta Project Manager). Diana Kukić Dasko (Lloyds Digital Business Development Director) steered the conversation.

Make sure AI tools add value to your business

AI tools are already making a big difference for businesses, especially in content personalization. And it is not just about text and images; it is also about personalizing the user journey, Mrvoš explained.

- In our main solution, Digital Experience Platform, the central platform for content in marketing and sales, we're currently piloting automatic translation of website content. With AI tools, you can translate your content into multiple languages and personalize it — all in the same place, and very quickly.

- Such tools bring value to companies by shortening the time from when a customer searches for what they want to the moment of purchase. With so many tools available on the market, it's important for us as partners to help our clients choose and implement them wisely, maximizing benefits for their business.

Expectations: The biggest challenge in implementing AI

It was interesting to hear how in digital agencies, besides technical challenges, the implementation of AI is influenced by one non-technical challenge.

According to Rajković, this is the expectation that AI can solve everything. - AI is powerful, but it requires a lot of knowledge and work to implement properly. On the other hand, we often fall into the trap where we have technology and tools, and then we seek problems to solve with them. I think that a needs-driven approach is much better - start by identifying the need and then find the tool to address it.

He stressed strong market dynamics, with tools that are constantly changing and new ones being launched every day. - It is challenging to stay updated and navigate that space. Moreover, you must be aware of the price. At CMG Digital we use about 10 AI tools, because there is no single tool that solves every requirement. When you add up the total price, the question is whether they solve the needs in a cost-effective way, or some kind of hype has been created around the whole thing.

How to protect sensitive data in interaction with AI?

For more than 20 years, Infosit has been a partner of Microsoft, which is known for its high security standards.

- It is very important to us to know where the data is located, said Mrvoš. For example, MS Copilot which is fully integrated into your system can help you prepare a report as it has access to your ERP. Previously, you needed an assistant for that, now AI can handle it. When it has a broader view of information, it provides higher-quality output.

Rajković pointed out that using sensitive data in interaction with AI is a challenge they face every day.

- Neither we, nor our clients, are prone to send sensitive data to a commercial AI model. Anonymization or abstraction of data could be an alternative, but again you are left with the issue of how to control who inputs what data. A strong awareness of what is confidential data and how to protect it is essential.

ChatGPT: Content quality and critical thinking

ChatGPT seems to be often used in both the business world and in education. Speakers agreed over the key concern related to ChatGPT: the quality of the content it generates.

According to Rajković, it is not necessarily accurate or the best possible. – Generative AI tends to hallucinate and distort data. It is trained on a large dataset and works on the basis of probability. It predicts what content it should generate based on our prompt. The quality of the output depends on the quality of the input (garbage in, garbage out). It is important to develop two essential skills. One is prompt engineering, aka writing good prompts. Another is critical thinking.

Miličić believes that in education, it is imperative to teach students to think critically so that they can evaluate results they get from genAI. Oftentimes, the output will be questionable and extra effort will be needed to make it better. He referred to the recent piece of research by an analytic company. - It showed that their consultants solve easier problems much better with the help of ChatGPT. On more difficult problems, the results were questionable. The greater the expertise, the greater the ability to solve problems, he concluded.

What does the future hold?

Although artificial intelligence has been used for decades, particularly in military technology, it took a long time to reach commercial application, Brenko noted.

- Lately, AI has been growing exponentially. Technology is following its progress, and it’s reasonable to expect the prices of AI-based solutions to drop. We already saw that with new versions of ChatGPT. We also saw that, fearing that OpenAI might take over the market, their rivals launched high-quality models to the market for free. Some of these models are even better in some segments, and can be used free of charge. For sure there will be a lot of competition in terms of models and solutions, he said.

He stressed the need for caution, as well as the need to train people in what AI means and how to recognize it, as it can be used equally for good and bad purposes.

Can AI replace us?

Is it realistic that AI will soon replace us? The panelists agree that such a scenario is not likely in the near future, but changes in the workflows and businesses are already happening.

Mrvoš thinks that AI tools will replace us in jobs that we don't even want to do. - They already help programmers with things they don't want to do, and in the future, they will help them even more. Developers have to change accordingly. And human change is slower than the progress of AI. We must not forget the change of companies, either. For example, if you are in online sales, you should expect your future buyers to ask AI ​​assistants on their mobile phones to find them the most appropriate shopping website. The website itself will no longer be a sales tool. It won't happen instantly, just like radio and TV didn't disappear overnight, but their impact will be reduced. Businesses should reevaluate their existing solutions as soon as possible and check if they still effectively address problems or if they need to look for new approaches.

Rajković believes that AI will not replace programmers soon, but the job description of programmers could change a lot. - At this moment, we can roughly see in which direction it could go, but it is not completely clear yet. The expertise for that position will likely rise. The focus will no longer be on coding but on other aspects of the role.

- Producing code with generative AI is practical. However, that code is what is called technical debt. It's still up to humans to analyze the bugs and code quality. We can add a layer of AI on top of that, but at some point there's a question of control over it. There will certainly be room for high-level expertise. Can AI replace the developers? It can replace some parts of the junior programmers' work. In my opinion, junior programmers are in bigger problems than programmers in general, says Miličić.

Vujičić believes that in most cases myths like this stem from fears. - These fears are like those we went through when we started using our first computers. If you apply AI correctly, in partnership with the market and your users' needs, then you can start new projects freely and boldly. They will result in better products and services.

He adds that AI has two major shortcomings and the question is when they will be fixed. It has no creativity and no intuition. - As long as we have ideas, creation, imagination, intuition - we will be irreplaceable.