I think AI will one day be better at evaluating some decisions than humans, says our colleague Daniel

Author: Zuzana Kocáková

Digitization

09. 02. 2024 10 reading minutes

At msg life Slovakia, we strive to keep up with innovation. An example of this is our two latest projects, where a job seeker can talk to an AI-equipped chatbot instead of reading a lot of text. The implementation of ChatGPT on the web is already yielding first positive results and the chatbot is currently being debugged. Behind this idea is our colleague Daniel from marketing.

You focus on online marketing, you have also worked in similar posts before. Why did you become more interested in artificial intelligence and its applications?

I was tired of writing reports. Each month I tracked selected columns and reported the data to people who mostly didn’t even look at it. So I tried to make this monotonous work easier for myself. I made a course through Udemy, using which I created an auto clicker. Its job was to regularly export data from the sites. I started to simplify the stereotypical part of my work and what could be automated, I automated. The only last step left for me was to check the data. Nowadays, similar parsing bots are common, which was not the case a few years ago. I also became interested in automating work.

To give a global reason, I think that each of us lives in their own ecosystem, equipped with different experiences, different skills, and thus sees only a slice of reality. Therefore, in my opinion, one is not able to judge things in a completely objective and comprehensive way. Humans and artificial intelligence (AI) have different ways of perceiving reality. We often rely on subjective perceptions, experiences, assumptions, guesses and emotions, whereas AI is based on the data and algorithms that are provided to it, thereby assuming a more objective outcome. That’s why I think artificial intelligence will one day be better able to evaluate some decisions than a human. (Read our article What is artificial intelligence? Is AI really intelligent?)

However, even AI has its limitations. For example, AI cannot recognize and interpret certain kinds of information that are obvious to humans. AI can also be affected by lack of data or poor quality of data it receives. Ultimately, both humans and AI have their advantages and disadvantages. Ideally, humans and AI could work together and leverage their strengths for optimal results. That is why I think it is already worth paying attention to the development of artificial intelligence so that we can exploit the strengths of both sides.

Do you have enough support from msg life Slovakia to implement similar innovative projects?

Our management believes that we cannot miss the trend and need to constantly explore new opportunities. We’ll see what it brings, but it’s always better to try than to be left behind. We also see the hiring projects Java developer and Actuarial developer as a learning process. We didn’t set one goal that when achieved, the project is over. We’re exploring ways to use AI and we’re always improving our websites. Based on our findings, we are also considering how to use artificial intelligence internally, how to use it to speed up some monotonous processes and leave employees more time for work where a personal approach is required.

Personally, I’d like to try connecting the chatbot to an email and experiment how it will respond. It should be trained on me and plausibly simulate what I would write myself. It takes me a lot of time to create an email and AI could do it faster. Artificial intelligence could also be used for the creative part of work, when we often wonder if the visuals are aesthetically pleasing. The AI should be able to create the most appropriate graphics based on the previous data. I have also recently been attracted to the idea of financial planning using artificial intelligence. What takes a human several hours to evaluate, an AI could do in seconds. The human’s job is then just to check the results and improve the algorithms. I have even read about a Polish company called Dictador whose management has been fully replaced by AI and so far they are generating good results. We will see how they continue to do. This is quite a rarity, so I’m curious to see how it all turns out!

We currently have a chatbot deployed on two sites. Its job is to answer questions to potential job candidates. What are the first results?

The vast majority of projects have very similar phases and it looks like this time will be no different. The first is to start the project with some expectations. If we were to imagine a graph, from this point the curve would rise steeply upwards in time until it reached a point called the optimism peak. This is the stage where imagination meets reality and often crashes into it. From this point onwards, the progress of the project usually has a rather decadent phase, followed by a phase of real implementation of the first requirements.

This is when the progress of the project hits rock bottom, which is why this phase is also called the Valley of Despair. But if the team overcomes it and solves all the obstacles, there is progress. Not as steep as at the beginning, when expectations were high, but a very “healthy” climb that corresponds to reality. Eventually the project reaches the Enlightenment phase, from which it continues to progress, but at a much slower pace, so much so that we could call it stagnation. This is because many lose their initial enthusiasm at this stage and it is the job of management to keep the project moving forward.

Our websites have already passed the initial stages and are starting to show the first results. They are visited by relevant people, now it’s up to us to decide how to use this data and how to refine the project to perfection. I see room to automate and speed up some parts. This would leave our rather busy HR department with more or less only the final phase, where they would connect with the suitable candidate and get a feel for their personality. The initial formalities and data evaluation could be handled by AI.

Both chatbots work using ChatGPT. Currently written communication is possible with them. What are the next steps in this project?

Since we don’t develop our own AI, we only use what Open AI offers. They have recently introduced a new feature, voice communication, and these days they are also launching ChatGPT Vision. Communicating by voice will be much easier, as one will no longer even need to click on the keyboard. Vision, in turn, is supposed to provide more personalized responses based on the emotions of the recipient. The ability to create a face for the chatbot is encouraged here as well. It’s up to the discretion of the management how an avatar representing msg life Slovakia should look like, I personally would go without the AI visualization and I have two reasons for that.

The first is that an avatar can generate (sometimes too much) sympathy, but equally dislike. Conversations could thus be significantly affected by this, which would unnecessarily undermine their factuality. The second reason is that the choice of avatar is not always a happy one. We see this with some humanoids. As soon as they have a face whose facial expressions are not always appealing, a kind of fear or even disgust arises, which is another negative consequence that we do not want to arouse. If we take inspiration from science fiction works, artificial intelligence is usually depicted virtually, as a kind of software that controls everything, but nobody can see it. We don’t want to create another human, we want to create something more intelligent. For this reason, our motif for the time being is the sphere. It is also an element that msg life Slovakia has in its logo.

At the moment, we have only one challenge left to solve, and that is the name of artificial intelligence. Most bots use the female gender. This is probably because the female voice is more credible, it inspires a kind of confidence. There are probably psychological reasons for it. A man is more likely to confide in a woman because they expect more understanding and support from her. I think this is true in the case of woman-to-woman communication, as well as man-to-woman communication.

What do you think are the shortcomings of ChatGPT so far?

Despite what a groundbreaking and very useful tool it is, we encounter quite trivial problems when working with it, such as the fact that the system crashes frequently. I’ve also noticed that if I retrieve data one day that the AI understands and generates a very relevant response, the next day it doesn’t remember this for some unknown reason. Apparently, such nuances are also caused by constant updates and improvements, which may not always improve the system. I am also limited in the number of experiments. After some time I get a message that I have filled my daily quota of requests and I have to wait. It is conditioned by the capacities that the system has and its creators probably did not expect such an influx of users. Another long-standing problem has been so-called hallucination, where the system’s priority has been to generate an answer at any cost, even at the expense of the truth. Such an excessive effort to informatively satisfy the user quickly crashed and the developers tried to remove it as soon as possible. I would also consider the uproar that is being created around the company to be a negative. It has to do with just where the company got the input from, and many of the facts lead to the conclusion that it wasn’t entirely legal.

On the other hand, it has to be said that compared to competing tools, ChatGPT has a big head start. But if I had to predict the future, I predict that Google’s Bard will catch up very quickly. It is not yet as well developed as ChatGPT, but it has the advantage that it processes real data. While ChatGPT has stored data up to a certain year, Bard can search through current information. This is, and I think will be in the future, an increasingly important difference. As for Microsoft’s Bing-Chat, for example, I think it generates the most appropriate answers. However, if all these companies were to come together, they would make a lot of progress. In my opinion, competition comes at the expense of quality in this case. They would make faster progress by working together, but they would lose the market leadership and dominance they seek.

How far do you think the possibility of AI in the world of hiring extends?

I think we can expect to see a lot of fake HR profiles on LinkedIn in the near future. This will also increase the number of spam messages. However, companies will have to do it in a subtle way so that messaging is not too fast and they are not blocked. As I mentioned earlier, I think HR will lose some of its workload. I wouldn’t see this as a negative, because firstly they will probably get rid of stereotypical work activities such as sending GDPRs, writing back to initial emails or searching and evaluating suitable candidates. For this reason, I think that recruitment agencies, whose role is to supply suitable candidates for positions, will also be at risk in the future. This will also reduce costs for companies that are currently investing significant amounts of money in the recruitment process. Investing in the development of AI or tools that use it can therefore be a profitable and rewarding investment for companies.

Do you think the staffing industry will be a lot less personal if the initial stages of recruitment are taken over by AI?

It will, but I don’t think that’s a bad thing. To give you an example: we used to buy a ticket for public transport from a lady sitting at a kiosk, but nowadays we do it with a machine or an app. Although we lost a short small talk on the way to work and the salesperson had to be retrained for another job, the service to users was improved. They can buy a ticket faster and can pay by card, which may not have always been possible before. It has improved the service to the customer and that, I think, is the goal of any company.

Also, if using chatbots reduces some of the emotions at the first interview, it may not be a bad thing. If a candidate (or even a recruiter) is having a bad day, this may not be taken into account and only the facts, i.e. experience, skills or expectations, will be considered in the first rounds. So the evaluation of whether a person fits into the collective would fall to the later round, which would be led by real people.

You’ve been working on AI for a while now, studying a lot about it and testing it from many angles. Is there anything about artificial intelligence that surprised you? Did you have different expectations or ideas than you have about its functioning today?

Yes, artificial intelligence has interested me for some time. I follow its development, research and try to understand its principles, functioning and how to make it work. In many ways the AI surprised me, but especially in the following areas:

• Speed of development: AI is evolving at an incredible speed. A few years ago, AI systems were only able to perform simple tasks such as playing games or image recognition. Today, they are capable of performing much more complex tasks, such as generating creative content or writing different kinds of text.
• Scope: AI is used in many different fields today, from healthcare and finance to manufacturing and logistics. Its potential is huge and expanding all the time.
• Ethical issues: AI also raises a number of ethical issues, such as the question of accountability for the actions of AI systems or the question of the potential discriminatory effects of AI systems. These issues need to be addressed in order for AI to be used safely and responsibly.

In the past, I thought AI would function more like a simple machine that can perform tasks as instructed. Today, however, I know that AI is much more than that. It is able to learn, adapt and can generate creative and original ideas. This surprised me a lot and I’m fascinated to see what AI can do in the future.

I believe AI has the potential to change the world for the better. It can help us tackle some of the world’s most pressing problems, such as climate change or poverty. However, it is important that AI is used responsibly and ethically. We need to make sure that AI is used for good things.

Share article

I think AI will one day be better at evaluating some decisions than humans, says our colleague Daniel

Zuzana Kocáková

At msg life Slovakia I take care of the company culture, events and I am part of the marketing team. I like a job where I can be creative and contribute to making my colleagues at msg feel good. I regularly prepare for you not only news and interesting things from the world of information technology and insurance, but also from behind the scenes events at msg life Slovakia.