On 16.01.2023, the hosts of Millásreggeli continued with Levente Szabados where they left off in December 2022. Levente is an artificial intelligence expert, co-founder and senior advisor of Neuron Solutions, and for the seventh time he had a discussion with the presenters about the past, present and future of artificial intelligence. This time, the future was the focus. Focusing on predictions for the year that has just begun, but also mentioning the year 2024, the discussion was once again very interesting, and we give you a glimpse of the details here.
Instead of New Year’s resolutions, Levente has compiled a list of what will happen in the new year in terms of artificial intelligence milestones, and what should, but probably won’t. But here the presenters’ imaginations immediately soared into less happy territory: “And what are the ones that shouldn’t happen but will?” “Well, unfortunately there are the latter, but let’s start with the more peaceful things,” Levente replied.
So what are the events that are likely to happen?
In general, the natural course of the diffusion of new technologies is that first in basic research (you probably only hear about it on TV), then in applied research – you may even see it at conferences. Then it reaches the development phase, when we are being pestered by our bosses, and finally the technology becomes embedded and we don’t even notice its presence, reaching the final, full-scale deployment phase.
Levente believes that in 2023, we are halfway between the development and final stages in many areas of artificial intelligence. Perhaps by 2024, the presence of AI will become natural. So it is likely that by next year we will find talking to machines natural. And by this, of course, Levente did not mean a prayer or a scolding to a broken washing machine…
We can expect to see a lot of conversational interfaces (human-machine communication platforms), customer service chatbots and the like, but we might already not be surprised. What will be revolutionary is the use of artificial intelligence to gather information. “Currently, if we want to know something, we use a mental paradigm called search, and we search.” We’re not asking Google a question, we’re trying to find the right keyword. That will change significantly.
But of course it’s not just the way we get information that will change, but the way we create information – are we still going to write marketing texts or are we going to generate all the content? There are already many such services available.
So there is no question that these new technologies will spread and become part of our everyday lives. The big question will be whether the technology will again be dominated by a few prominent players in the market whose services will be used as a magic black box by everyone, or whether, in the case of Stable Diffusion – which is an open source AI model that generates images from text – the code will spread and anyone can create a service, even on their own machine. This is the big question of 2023, which thinking will win.
“Is it a battle of mindsets or a battle of financial economic power?” – “Both.” It will depend on whether it is possible to build a new type of AI model without big money. Right now we know that if a model is big and expensive enough, it works. That doesn’t mean we know it doesn’t work on a small scale, it just means we’ve only managed to do it on a large scale so far. If we could do it on a small scale, then widespread take-up would happen. But if it turns out that it can only be done at this scale, because the laws of nature say so, then that is the unpleasant situation where this kind of monopolisation almost naturally sets in.
Well, after that little digression, back to Levente’s list!
Automation in industrial applications was the first on the “could happen, but won’t” list. According to Levente, the huge amount of sensor data available should be used to predict failures and prevent scrap. Of course, this is already underway, but it is moving at a very slow pace.
Besides, self-driving cars probably belong on this list. “I have a suspicious feeling that it’s not the technology that’s holding us back, but when we can get the social consensus to go for it.” – well, if Levene has that suspicious feeling, there must be something to it…
And speaking of social consensus, it’s worth adding to this list the question of what we do to people if we automate a process completely. Perhaps this is one of the main aspects that slows down industrial automation in this way.
So we have seen that there are plenty of things that need to be addressed, but it seems that we are not going to do so for the time being. Well, unfortunately, there are also things that we shouldn’t, but almost certainly will. Such activities include, for example, the generation of propaganda content or fake news, or the production of deepfake videos. “So it’s almost certain that sooner or later we’re going to have to get used to the idea that what we see is either something that we’re going to be terribly critical of as humans, or we’re going to have to specifically use another machine and service that’s going to critique for us whether what we’re seeing is realistic.” – Wow, that’s not a very peaceful vision.
But to end the discussion on a lighter note, Levente was also asked about the development of artificial intelligence in Hungarian, and he had some good news to share. On the one hand, we can already see for ourselves by trying out ChatGPT that a very high quality of Hungarian conversation has already been achieved. On the other hand, there is an ambition to make larger training texts available to everyone soon: a corpus (a collection of texts that can be used to teach AI models) of several billion words is expected to be released, which anyone will be able to use to teach. So Levente is optimistic in this regard.
Well, then, we are looking forward to the next discussion with great optimism, because the rapid development of artificial intelligence will ensure that there will be something to talk about with Levente at the next Millásreggeli too. For more details, listen to the conversation in Hungarian at this link.
And don’t miss out, follow us on Facebook and LinkedIn!