What do the movies Her (2013), Terminator 3 (2003) and Bicentennial Man (1999) have in common? They all feature machines that have the ability to mimic human intelligence or in other words Artificial Intelligence (AI). And at some point, you might have wondered, “Is a sentient AI a possibility?”
AI has been in the picture since the late 1950s and the topic of sentience has always been a matter of debate. But it was established long back that AI lacked the capacity to emote, so why is the internet suddenly buzzing with the talk of sentient AI? Before we get into the ‘why’ of things, let’s have a little backstory with the ‘what’ and ‘how’ of it.
What is AI?
According to Google, “AI is a set of technologies that enable computers to perform a variety of advanced functions, including the ability to see, understand and translate spoken and written language, analyze data, make recommendations and more”.
AI came into existence around the 1940s and 1950s when scientists, mathematicians and scholars were playing around with the idea of creating an artificial brain. It was in 1956 that artificial intelligence became an academic discipline under Alan Turing. Prior to that, computers were machines that could solve problems but lacked the capacity to remember what they had solved earlier.
Since then, AI has come a long way. We have cars that can be driven around driverless, we have phones that unlock with facial recognition, and we have chatbots writing entire essays for us. AI has spread its roots in all sectors from banking to healthcare to manufacturing. But this is just the beginning of AI. AI has 3 stages of evolution which are:
1. Artificial Narrow Intelligence ANI: The first stage of AI development that we are currently in. In this stage, artificial intelligence is able to recognize text and images, analyze data and produce results based on it.
2. Artificial General Intelligence AGI: This is the second stage of AI development when AI attains human-level cognitive abilities in fields like language processing, image processing and reasoning.
3. Artificial Super Intelligence ASI: This is the last stage of AI development when the level of intelligence of computers crosses that of human beings.
Today, AI has progressed to such a level that it stands on the threshold of AGI. While some people might be ecstatic to hear this news, seeing endless possibilities, there are some to whom this news is alarming. After all, there must be a reason why AI experts like Elon Musk, CEO of Tesla, Steve Wozniak, Co-Founder of Apple and Emad Mostaque, CEO of Stability AI have signed an open letter, requesting companies to halt the further development of AI.
How far has AI come?
While mulling over the idea of machines being intelligent, Alan Turing had come up with a test to determine intelligence. If an AI passed the Turing test, it could be deemed as intelligent as a human being. And a lot of AI today can pass the Turing test. Deep Blue, IBM’s supercomputer defeating the world chess champion in his game in 1997 is the perfect example of a machine surpassing human intelligence.
In recent years, there has been a rise in the human-like trait that AI is capable of pulling off. An example of a machine showing human-like intelligence is ChatGPT, an AI language model developed by OpenAI. An experiment was conducted by researchers at Microsoft, where they asked GPT-4 how to stack up a book, 9 eggs, a laptop, a bottle and a pin, and ChatGPT’s answer was jaw-dropping. Not only did the chatbot provide the right order of stacking up the objects, i.e. the book, the eggs the laptop, the bottle and finally the pin, but it also provided added instructions on how and why they had to be stacked in that order.
ChatGPT instructed that the eggs should be placed in three rows of three, with the laptop on it, with the keyboard facing up and the screen facing down. It further explained that the laptop will act as a solid base for the bottle, which should be placed empty and should not be shaken. Atop which the pin should be placed which would have to be clean and the pointy edge should be on top. This answer not only proved ChatGPT’s ability to solve complex problems, it also showed its awareness of the physical nature of the objects. This led to Microsoft releasing the study stating GPT-4 displayed ‘sparks of AGI’.
Does this mean that AI has reached the level of sentience? Is it capable of feelings? Is this the reason why Sam Altman, the creator of ChatGPT said he is “a little bit scared” of it? To answer that, we need to understand what sentient AI is.
What is Sentient AI?
For artificial intelligence to become sentient AI, it needs to have the ability to feel and judge its environment and make decisions like human beings. Simply put, AI should be intelligent enough to be able to think like humans. So, can AI do that?
What makes human beings, as well as some animals sentient is their ability to be aware and make decisions subjectively. (Dogs have evolved to have facial muscles that are missing in their wolf ancestors which they use to look guilty/adorable when they are being scolded). However, we do not know the source of this sentience. Science is aware of where the brain is but is not aware of where the consciousness is. And this is where the complications begin.
As Sundar Pichai, CEO of Google said, “There is an aspect of this which all of us in the field call as ‘black box’. You don’t fully understand, and you can’t tell why it said this, or got wrong. Let me put it this way. I don’t think we fully understand how the human mind works either.”
How close are we to Sentient AI?
While Microsoft did point in the direction of AGI, it did not claim to have built a sentient AI – something that can prove to have a disastrous effect on the careers of scientists and engineers. Blake Lemoine, Google’s engineer who claimed AI was sentient was fired. According to sources, Google’s LaMDA AI was not only able to hold deep conversations, but also stated that “The nature of my consciousness/sentience is that I am aware of my existence, I desire to learn more about the world, and I feel happy or sad at times.”
How far away are we from Sentient AI?
While the conversation certainly seems to resemble something, a human being would say, but it is also important to note that language models are built and trained to have conversations that resemble human interaction. Human beings also have the tendency to attribute human qualities to non-human things. Something that is also known as Anthropomorphism.
While language models and today’s AI definitely show signs of intelligence and the LaMDA AI did pass the Turing test, they still need a lot more training to reach AGI, let alone sentience. For example-apart from logical thinking, AI is also judged on its linguistic capacities and its ability to translate languages and form coherent sentences. Remember the hilarious translations Facebook used to provide for text in foreign languages? Now there are barely any noticeable mistakes.
Translated’s CEO Marco Trombetti explained, “Language comes naturally to human beings, but the same cannot be said for AI.” According to a report, The amount of time an editor takes to fix mistakes made by a human translator is 1 sec/ word, the same editor takes 2.5 sec/word for text translated by AI. What is interesting to note is just 10 years ago this number was 3 sec/word. The difference might seem marginal to you, but it is quite big in the AI field and also shows a definite improvement.
So, if sentience is still not on the horizon, then what was the reason that big companies around the world petitioned for the halting of AI development for six months?
Not all scientific discoveries are safe for human beings – an example of which is the Dolly Experiment where a sheep was cloned. While AI is a powerful tool capable of amazing feats, it is also something that should be approached with caution. Even though AI technology has existed since the 1950s, its development in recent years has been exponential. The technology is now advanced enough that it is being used to fool people and can be misused.
There is also an issue of limitations. Human intelligence is limited and only capable of certain levels of intelligence, that is not the case for AI. As long as there is enough space and power, AI levels of intelligence can keep growing.
Yet the worst possibility is Artificial Super Intelligence. Calculations suggest that once AI surpasses human intelligence, it will be beyond the control of human beings. Since human beings will lack the intelligence to keep it under control.
“One can imagine such technology outsmarting financial markets, out-inventing human researchers, out-manipulating human leaders, and developing weapons we cannot even understand. Whereas the short-term impact of AI depends on who controls it, the long-term impact depends on whether it can be controlled at all.”
When Google’s conversational AI Bard AI was asked “Do you have feelings?” It was quick to respond that it lacked the physicality to experience emotions the way humans do, but it still felt joy in helping people out. While the answer is endearing, Artificial Intelligence is also proving to be a double-edged sword. Technology has already brought into existence Self-Teaching AI. It will be interesting to see the effect it has on the world.
What are your thoughts? Do let us know in the comments.
Disclaimer: The content above is solely the author’s opinion and should not be used to influence any party. The author and the Wire19 team are not liable for any investment or financial decisions that may be made on the basis of this article.