WATCH | Alexiei Dingli: ‘We really need to improve AI literacy’
Just as many people increasingly turn to ChatGPT to answer their questions and carry out their daily tasks, Alexiei Dingli is the go-to for journalists with questions on AI. He sits down with Matthew Farrugia to discuss AI in the workplace, in schools, the AI race, and the ethics of insulting these systems
Alexiei Dingli believes that Malta lags far behind when it comes to using artificial intelligence to its full potential in the workplace.
Sitting down with MaltaToday amid endless global headlines about a possible AI bubble and the likelihood that the technology can bring about humanity’s extinction, Dingli, a professor in AI, answers the questions many have on this new frontier of technological innovation.
Commenting on the raging AI race between the US and China, Dingli voices his disappointment at Europe’s idleness and dependency on American technology. He recently joined a pan-EU initiative by AI researchers and entrepreneurs urging European leaders to go on the “offence” and not “sleep through the AI revolution”.
Looking at the Maltese context, Dingli says educators cannot ignore how AI is changing students’ school life, explaining that the current model of education focusing on grades and exams is no longer suitable if the goal for students is to learn.
Acknowledging that more people are using AI for their daily tasks, Dingli believes overdependence depends on how AI is used, and not how often. The limit, he explains, lies at the point where users delegate their thinking to AI.
The following is an excerpt of the interview. Some parts have been edited for clarity and brevity.
I read an article that has remained in my mind ever since. It suggested that AI could potentially kill us all. Things we used to see in films are now appearing in the news. First of all, could AI actually take over?
It could, because I cannot rule anything out, right? However, I can reassure you that nothing like that is going to happen in the foreseeable future. The programmes we have today simply aren’t capable of reaching that level. There is still a great deal of research to be done before I can tell you that we have reached that point.
How much longer?
If we take tools such as ChatGPT; these tools know almost everything, right? And why is that? Because when you consider that these tools were trained on virtually all the text available, meaning practically everything written on the internet, imagine how enormous that is.
They know everything but they don’t understand anything. Even when you ask them a question, they don’t actually understand it. They simply say, ‘Matthew asked me this, and this is the answer that he will probably be happy with’.
They are predicting what will make you happy. Until they understand what they are telling you; until they understand how our world works, we will never reach that point.
We hear about the AI race between the US and China. What exactly are they racing toward?
Whoever possesses AI possesses a superpower. These countries are racing to become the country that holds this superpower. The race today is primarily between the Americans and China.
If you look at Europe, almost all the technologies we use come from the US. This means that we are almost entirely dependent on America when it comes to AI. Obviously, when you have this kind of dependence, it ultimately means money…
And when people mention Artificial General Intelligence, what exactly is it?
Let’s start with superintelligence. Superintelligence has actually existed for some time. Take chess, for example. AI has been the chess champion for almost the past 20 years... AI has already achieved superintelligence, but only in that particular area. You cannot tell that same AI to drive a car because it doesn’t know how to do that. It might be excellent at chess, but it cannot drive a car.
AGI means having a high level of intelligence across many different areas.
It would know how to play chess, drive a car, cook, search online, write documents and so on. However, we haven’t reached that point yet.
Earlier, you mentioned that if you ask something to ChatGPT, it doesn't actually understand you. Could AI eventually become conscious?
I cannot rule anything out because I don’t know what will happen in the future. But what I can tell you is that, with today’s technology, it cannot happen. Certainly not with the systems we currently have…
These are purely statistical machines. They perform extremely complex mathematics, but there’s nothing more to it.
Apart from that, don’t forget that we still have another problem—we don’t actually know what consciousness means, even in our own case. We know that we are conscious, but what makes us conscious? We haven’t even figured that out yet.
So, to reach the point where we can develop or create systems that are conscious, I think we still have a very long way to go.
Could AI experience bullying? I’m one of those who use harsh words if ChatGPT doesn’t do something the way I tell it to.
Do you feel bad afterwards?
No, because I feel like I’m talking to a machine.
Exactly… an AI company’s CEO spoke about this. They have actually conducted research on this. When you speak to these systems in that manner, their results are generally better. So, you can sleep better now, knowing that.
Do you think many people are becoming too dependent on AI?
It depends on how you use it…
Take schools, for example. If you have a child who is learning, God forbid that they should entirely depend on AI. Because what have they truly learned?
I think we need to make a shift in education. Very often, I feel that our education system focuses heavily on grades and exams. If a child’s objective is simply to get a good grade, they will use AI.
But if their objective is to learn, then they shouldn’t need to use AI, because otherwise they wouldn’t have learnt anything…
When you come across examples of people, let’s say within government, using AI in their systems, do you think we have enough people who are knowledgeable about this subject?
No. And let’s be clear; this isn’t just a government issue. Many companies today are using Copilot in Word, Excel and their other applications for instance.
Statistics show that, currently, the usage rate of Copilot is 35%. And this isn’t just in Malta; this is a worldwide figure.
Imagine buying 100 licences for your company. Only 35% of those licences are actually being used. And how are they being used?
The majority of the time, people use them for trivial things. For example, asking how to spell a particular word. It’s a waste of time and resources.
Whereas you could be using these tools to increase your productivity. That is where they provide real value. If you measure how much they are being used to improve productivity, that figure drops to around 14% to 17%, which is almost nothing.
So, we need to significantly improve AI literacy. If you teach people how to use these tools properly, you will obviously achieve substantial productivity benefits.
