Can AI Replicate Human Behaviour?

Since artificial intelligence (AI) technology has become rapidly more advanced, seeing significant improvements in recent years in particular, questions about its potential are becoming prevalent.

The primary concern among professionals across just about all industries pertains to the ways in which AI may be able to perform their jobs better than them.

However, another area of great intrigue is the ability of advanced AI to behave like humans on a deeper level. The idea is that since AI is now able to do things that could only be done by people in the past, how far will its capabilities go?

Of course, we know that it already has the ability to do things like write complex texts, recognise speech and so much more. So the burning question is, if AI can takeover all these tasks from humans and do them just as well, if not better, can AI behave the way humans do too?

Most of all, we’re talking about expressing and perceiving emotion – detecting, for instance, when somebody is sad and providing an empathetic response. Or, on the other hand, being able to be funny.

The most basic answer is that AI technology can’t actually feel human emotion, but that doesn’t mean it can’t be trained to mimic it. The question is, how believable is it, and will we be able to tell the difference between human emotion and AI-manufactured emotion?

 

How Well Can AI Mimic Human Emotion?

 

While a lot of the fear that used to surround AI has been quelled as people have improved their understanding of AI and what it can and can’t do, the apocalyptic notion of “robots taking over the world” still seems to persist, even if it’s just mild background concern.

At the core of this fear is our ability to be able to separate humans from AI, and that fear isn’t altogether unfounded. In fact, there are many scenarios in which this has already become difficult.

Just think of bots on social media and the need for the proliferation of CAPTCHA technology. Who would’ve thought we’d live in a time when it’s normal to be asked whether or not you’re a robot on a daily basis?

This issue, however, is still distinct from the idea of replicating human nature. While we can say, without a doubt, that AI is advanced enough to trick us into believing that work online has been done by a human, that’s still pretty far off the ability to convince people that they’re actually talking to a human as opposed to an AI.

Of course, we’re talking about going beyond merely setting up a profile on a social media platform and having a bot conduct conversations with real users. We know that while AI certainly can hold up a conversation, the general feeling is that it lacks a human element to it.

That is, emotion – both our greatest strength and, in many cases, our greatest weakness.

While AI obviously can’t actually feel emotions because it is, for intents and purposes, a robot, that doesn’t mean it can’t pretend to feel them. To be fair, it’s not like humans don’t do it from time to time, and often, get away with it.

So, how well can AI pretend? Can it actually mimic human emotion well enough to fool us into thinking we’re talking to a real person?

The answer, which is neither simple nor conclusive, is basically yes. Based on the current capabilities of AI and the way in which the technology works, it certainly seems fair to assume that as it continues to progress, AI will get better and better at convincingly mimicking human emotion.

The main reason for this is that AI technology is improved by means of training it, and that entails not only feeding it lots of information to learn from, but developing it in such a way that it just gets smarter and smarter every time it’s interacted with. That is, the more samples it receives that display emotions of different types, the more it’ll learn from these examples and start implementing these lessons into its behaviour.

However, while AI has the potential to get very good at mimicking emotions and pretending to feel things in the way humans do, there are still major limitations in this endeavour.

 

 

Can AI Be Funny?

 

The idea of AI having the ability to be funny really follows the same line as its ability to express human emotion. AI can’t actually generate unique humour because, obviously, it’s not human, but that doesn’t mean it can’t replicate humour.

By learning from humourous content used to teach AI programmes, it’s possible that they can be trained on the parameters of humour. That means including topics that are and aren’t considered funny, how to deliver a joke, the format of a joke and more.

However, just how funny the AI really has the potential to be is totally based on how it’s being taught and with what, and how much, content. AI has the ability to mimic humour, as it does with emotion, by learning from other examples of humour, but this means that it’s also limited by the material used to teach it.

Essentially, AI can’t generate unique humour or really be funny in a way that’s comparable to humans.

However, AI tech progresses and programmes are trained more and more and are provided with greater variations in samples, their ability to replicate humour will improve beyond merely telling jokes.

 

Limits to AI’s Ability to Mimic Emotion

 

By mimicking emotion and specific emotional responses, AI certainly can be convincing in attempts to pretend that the technology is human. However, the technology has its limitations.

First, AI needs to learn from the information that it’s provided with. This allows it to see what kind of responses are given in particular scenarios.

However, a complex aspect of emotional responses is that humans not only react emotionally, before we do that, we analyse the situation we’re responding to, whether that’s consciously or subconsciously.

As a result of that analysis, humans produce an emotional response, and that is what forms part of the roadblock for AI tech pretending to be human. That is, evaluating how we’re feeling, knowing when and how to react and doing so in a way that can be as unpredictable as humans.

Essentially, AI technology is limited by the fact that its ability to analyse situations is based purely on what it’s been taught by means of samples and more. This can lead to responses being predictable and limited.

However, having said that, progress in sentiment analysis – essentially, the ability of AI to evaluate human emotions – is coming along steadily, according to many experts in the field. This is conducted according to data and the evaluation thereof.

AI will never have an infinite selection of “emotional” responses, but it is impossible for the technology to be constantly improved by allowing it to learn from more samples, giving it the ability to have a broader range and, subsequently, improved authenticity which will help increase the accuracy and appropriateness of the responses.

For now, however, while AI tech continues to progress and improve in is ability to mimic emotion, it still has a way to go before it will, if ever, be impossible to detect.