AI chatbots are the new wonder of the internet, wowing audiences with their revolutionary abilities. The leap seems huge. It’s no small thing to chat with a chatbot „as if” you were chatting with a person and not be able to tell the difference without asking it direct questions about its capabilities.
But can chatbots think, like humans? Do they really understand what they’re saying, or are they just making connections that seem logical, or at least perfectly harmonized, between the enormous amounts of information they’ve been trained with?
If chatbots rely 100% on processed information and can’t make other connections, we should know what those connections are and how accurate the processed information is. Several major Western newspapers have been investigating what the training sources of the apps are and, not least, what the limitations of these AI apps are. What Romania wants to change in its Recovery and Resilience Plan? The Ministry for European Funds proposes vouchers for solar panels, a metro line to Otopeni and the elimination of the 9.4% of GDP
G4Media spoke to physicist Cristian Presură about the capabilities of chatbots and the possibility of them making the leap to more than „simply” harmonizing the huge amounts of information they have been trained with.
Can chatbots think like humans? Do they really understand what they say? Can they develop consciousness, and if so, when might such a leap occur?
„Personally, I don’t currently see a physical property that would act as a definitive barrier to artificial intelligence developing consciousness.
The physical basis of our consciousness and artificial intelligence ( chatbots for example) are atoms.
Bringing atoms together, for example, builds our bodies together with neurons. In these, at some point, consciousness emerges.
Looking at it this way, another organization of atoms I think could develop consciousness, if there are no natural barriers we don’t know about, or other necessary conditions than physical ones.
But if there are no such barriers or conditions, then the question is not if that transition will occur, but when.
But to get there will require a much more complex hardware architecture than what chatbots have now (compared to human neural networks), and a software architecture that comes much closer to the way our neurons work together (if we want to mimic human thought),” replied Cristian Presură.
He says he is amazed by the capabilities of chatbots using artificial intelligence, and that his expectations have been exceeded.
„Today we know a lot about our brain, but at the same time very little about the essential processes of consciousness.
For example, we don’t know how the simple thought of turning off the water tap is expressed within neurons.
How could we then implement this thought in artificial intelligence, or worse, if artificial intelligence…