For several months, ChatGPT, this tool equipped with high-performance artificial intelligence, has been at the heart of technological news.
Presented as a conversational agent, it allows users to dialogue on an infinite number of themes with a chatbot, based on the linguistic model of artificial intelligence GPT-3, developed by the company OpenAI since 2020 and launched more recently. ChatGPT is able to provide an impressive level of answers, to the point of bluffing more than one.
A revolution that upsets the perspectives of uses in the landscape of AI for some. The return of the wildest fantasies or worries sometimes considered irrational for others.
In the multitude of current debates, we interviewed two of our Minds experts: Jean-Christophe Bonis (futurist, entrepreneur and passionate about new technologies) and Rand Hindi (French entrepreneur and visionary in the new technologies sector) around two fundamental questions.
Does ChatGPT foreshadow an anthropological revolution?
Chat-GPT would be one of the last expressions of a great digital transformation at work that is based on quantification and expertise of the real. ChatGPT has the capacity to conduct conversations in a language similar to ours, based on its data and knowledge.
To what extent is ChatGPt pushing something social? Are we seeing a change in the status of digital technologies?
The use of artificial intelligence is positioning itself as a real springboard for businesses and individuals. It has the potential to transform and facilitate many aspects of our daily lives, which could lead to significant changes in the way we live and in our society.
However, it remains limited in its ability to understand and reproduce human intelligence autonomously. This new model of interpretation has the ability to manipulate language, but it cannot, unlike humans, convey vital momentum.
Ultimately, it is not us who learn to speak as a human being, but the human who speaks as a machine. As such, it is important to maintain an ethical and responsible dialogue about the uses and behaviours that ChatGPT may face.
The point of view of Jean-Christophe BONIS
In Barjavel’s ‘The Night of Times’, in the middle of the last century, the author depicted a universal translation tool allowing all multilingual members of a scientific mission to communicate in real time. In a way, he was the ancestor of ChatGPT from the fertile imagination of one of the best French science fiction writers.
More than half a century later, human intelligence created not a universal translator but a universal writing tool. On paper the promise is big, but the reality is much more nuanced…
At the risk of sounding like the devil’s advocate, the road remains very long. ChatGPT
is a perfect image of our current society, the hyper-consumption of content in McDO mode at the expense of relevance and quality. The current proposed use is a fundamental aberration for a very simple reason: ChatGPT has no idea what he’s saying! The present technology, although impressive in its ability to formulate a language close to man, is only a statistical representation of convincing human communication on form but without substance.
For the tool to really pass the technological gadget, we would have to have reached the level of an artificial intelligence capable of nourishing a reflection and an argument. The size of the model and the computational power required is so far unattainable in the state of our scientific knowledge. Tomorrow perhaps… and this would not necessarily be good news for man.
Lack of understanding leads to two major pitfalls: disinformation and
In other words, if I may say so, saying something stupid with elegance doesn’t make you any less stupid!
The baby is not, however, to be thrown out with the bath water. The potential of ChatGPT on many vertical looms is real. Its use for customer services will make our life so much less complicated when we call his bank, his insurance, social security or the post office. The concept of software development and computer code writing will never be the same again. Instant translation will finally become a reality that will forever abolish the language barrier.
In conclusion, like a weapon, ChatGPT can depending on its use do as much good as evil. Understanding its foundations, potential and limitations is critical.
Its use in the right place will strengthen you, any error of appreciation will destabilize you.
ChatGPT: dressable or uncontrollable?
There are many questions about the future of ChatGPT and the potential risks of its drifting and skidding to generate ideas such as training the tool to better control it.
Although ChatGPT is constantly being improved in terms of the quality and relevance of its responses, it may still contain errors and social or moral biases due to the data used to build it.
It is important to note that the Conversational Officer does not have the willingness and ability to make decisions independently.
Without being alarmist, it is essential to ensure that this type of tool is used in an ethical and responsible manner.
Rand HINDI‘s point of view
ChatGPT is an absolutely incredible technology and personally, I was the first surprised by how this technology works. We know that since 2017, the bigger the model, the better its performance. So we suspected that at some point it would eventually work, just by putting more computing power and more data.
However, just because the model is capable of copying human conversations does not mean that it is an autonomous agent capable of making its own decisions. Today, ChatGPT is still a tool, a tool to which we can ask incredible things and to which we can make very dangerous things, but which has absolutely no autonomy, no consciousness. And it’s really important to distinguish that. Just because we are able to predict and answer questions intelligibly does not necessarily mean that we are able to make decisions independently.
So, is it possible that someone is plugging in what’s on a system that would control something? Yes, of course, but I don’t think that’s the issue at all at this point.
Another point, to be able to achieve a super intelligence, it would take even more computing power. But we already see that at this time we are reaching saturation of the number of electronic chips available to run this AI model. So everyone is overexcited but it is not even clear that the digital infrastructure today even allows to support a super intelligence if we continue in the same AI as today. So I think that in order to achieve a super intelligence, we should first make it autonomous, which is not the case today.
Secondly, it would take models 10 to 100 times, see 1000 times bigger than ChatGPT, which is not possible with current technologies because there are not enough computing powers available. So we might have to switch to quantum or photonics to do that.
And thirdly, the problem of consistency of answers should in fact be solved. That is to say that today, we are not able to differentiate when ChatGPT tells anything, from when ChatGPT tells something correct. And so that, I think is a fundamental problem for adoption.
Absolute control and the replacement of man by AI are therefore still not on the agenda but the potential offered by ChatGPT on intellectual tasks remains nevertheless to be considered by companies, This includes new perspectives related to process automation and the simplification of complex tasks.
New robot or new buzz whose craze will fall much faster than we think? Jean-Gabriel Ganascia, Chairman of the CNRS Ethics Committee, denounces the «myth» of superhuman artificial intelligence and reminds us that AI software is not endowed with reason.