How long will it be before LLMs are indistinguishable from humans?

[Sparky_007]

I am sure this has been asked here in many different ways. But i will just ask again. Are we headed to AGI? Some experts believe that LLMs will one day reach a point where they are indistinguishable from humans. What are the implications of this for our society?

 

[svjiminfxv]

LLM's becoming indistinguishable is worrying, at the moment it is fairly easy to tell if an LLM has been used and it's not human created but as they develop, it will become more difficult. This will give many the chance to make their work easier by using LLM's without getting caught.

 

[Kali]

I don't think the lines that divide machines and humans will blur. there will be differences, and in fact, the differences will even become strong. LLMs are created by humans, they cannot surpass human intelligence, they cannot think creatively, they do not have originality, their cannot give new interpretations, they know what they have been told.

 

[Ja sa bong]

Is what you're asking possible? The answer is yes in my opinion but the real question is when or how long will it take before this stage can be reached? Unfortunately, there's no way of telling that for now.

 

[Heatman]

No, I don't think that it's possible for there to be no difference between human being and AI program especially when they're going to built into physical machine that's going to interact with man. Yes, the human evolution with AI is either set to be a world of human/machine collaboration or where human being are transformed into machines but the difference will always be clear.

 

[Ja sa bong]

LLM's becoming indistinguishable is worrying, at the moment it is fairly easy to tell if an LLM has been used and it's not human created but as they develop, it will become more difficult. This will give many the chance to make their work easier by using LLM's without getting caught.

If it's about what we are talking about being worrying, it surely is because a lot of things can go wrong and those who knows how to exploit such AI tools for negative use will take advantage of most things they can use it on.

 

[Kross]

It is not possible for language model contents to not be differentiated from humans. There are huge differentiations in how every human of a particular set and subset expresses itself with contents. The variation is so large that AI models won't contend them all. It would only be able to imitate popular demography which would still make them look mechanical with time.

 

[Heatman]

It is not possible for language model contents to not be differentiated from humans. There are huge differentiations in how every human of a particular set and subset expresses itself with contents. The variation is so large that AI models won't contend them all. It would only be able to imitate popular demography which would still make them look mechanical with time.

Yeah, that's very correct. Even if those AI language model can be able to master this, it's not something that's going to perfected in a short period of time. There are billions of data for it to process and have every information it needs to print it's own results to sound like human beings. It's a long term project that can never be rushed.

 

[Fidelia]

I think it would take ample time before LLM would get the near perfect disposition of humans in writing. Different humans express emotions differently in writing and it would take large and varying sets of data for a language model to incorporate all those emotional nuances into it for it to generate perfect human like writings.