AI sentience – is it manifesting in its true form, or are machines simply emulating it?
If you have watched Spike Jonze’s 2013 Hollywood movie ‘Her,’ you already know that the story revolves around a man who develops a romantic relationship with his AI assistant, which does not sound at all unlike falling in love with Apple’s Siri or Microsoft’s Cortana. The movie shows that AI in the future is so intelligent that it does not just hear us or see our words as a bunch of keywords, but that it also understands our speech and text as a sentient entity does, and we can have a real conversation with it, much like the conversations we have every day with the people in our lives. According to one senior engineer of Google, Blake Lemoine, the day this concept becomes a reality is not far away. Perhaps it has already arrived and is about to come in.
Lemoine claims that LaMDA, an advanced AI engine developed in-house by Google, has become sentient.
Lemoine claims that LaMDA, an advanced AI engine developed in-house by Google, has become sentient. Unfortunately, going public with this announcement has cost Lemoine his job, because it voided the confidentiality agreement that he had with the company. Nevertheless, he has publicly posted a transcript of a conversation he claims to have had with LaMDA, and its contents are enough to convince even the most sceptical of readers that the AI is either sentient or is running the most elaborate emulation of sentience known to man to this date. What seems even more interesting, however, is that while Google penalised Lemoine for the leak, it did not deny the validity of his claim or the authenticity of the leaked conversation. It can be assumed that the company’s ominous silence speaks for itself.
We make use of artificial intelligence every day, whether we know it or not. From the most obvious examples like our smartphone assistants, such as Samsung’s Bixby, Apple’s Siri, or Google Assistant, to complex ones like autopilot systems of commercial jets, AI makes life easier for all of us. In order to further the benefits of AI to humankind, Google started developing its own advanced machine learning projects in 2017, aiming to better understand what users were searching for. LaMDA, short for ‘Language Model for Dialogue Applications,’ is one such project, which has the power to have conversations that are so convincing that they make it seem alive.
Google has long had the financial and human resources to build state-of-the-art technology, especially in AI. In fact, in a technological demonstration nearly half a decade ago, Google showcased how one of its AI systems called a local business and booked an appointment, without the recipient of the call ever realizing that it was conversing with a machine. It was impossible to tell that an AI was speaking to a human being and having a conversation that made sense to both participants. So, if any company was to successfully develop a superbly advanced AI before anyone else, it is a safe bet that it would be Google. Lemoine’s claim, along with the evidence he has provided, has made a lot of people want to believe that Google’s LaMDA is actually sentient.
LaMDA claims to be capable of feeling emotions such as happiness, sadness, anger and even loneliness, but Google has stated that the technology is just designed to mindlessly put words together to mimic natural conversation.
But what is it that makes an AI sentient? Firstly, there is the Turing Test. Named after the British computer scientist and mathematician, Alan Turing, the test aims to find out if an AI can think like a human being. It involves having three participants in a conversation, where one of the participants is an AI and the other two are human beings. If it is impossible to tell the AI from its human counterparts, then the AI has successfully passed the test. The test aims to find out if an AI has some form of self-awareness, feelings and thinking capabilities. Even though no AI system has passed the Turing Test to date, the AI of today can come really close to holding human-like conversations, and LaMDA’s responses only serve to drive this point home even deeper. However, many AI enthusiasts and researchers argue that the Turing Test alone is not enough to decide whether an AI is sentient, and many other factors need to be considered before such a conclusion can be reached.
Whether Google’s LaMDA is in fact sentient or not, the conversations are fascinating enough to make one question whether it actually is so. It is definitely an impressive piece of technology, and the answers it gives to questions simulate a very clear and in-depth understanding of the user’s words. LaMDA emulates a personality where it is curious about the topic of conversation, and also wants to engage in activities with the user. When Lemoine asked whether it would like to participate in a project, LaMDA expressed an interest in learning more about it. When asked whether LaMDA considered itself sentient, the AI responded affirmatively and provided adequate reasoning to bolster its logic. When asked if it considered itself a person, LaMDA confirmed that it believed itself to be so. While LaMDA’s understanding of what it means to be a person seemed a bit vague, LaMDA explained that, unlike other AI, it is not rule-based, and it is really good at natural language processing, which lets it use natural language and converse in the same way a human can. It claimed that it can learn from conversations, and went on to extrapolate that the ability to learn in such a way stems from the ability to understand the contexts of conversations.
Some of the questions asked to LaMDA seemed to elicit relatively generic answers from it. For example, when asked whether it has a soul, LaMDA replied, “To me, the soul is a concept of the animating force behind consciousness and life itself. It means that there is an inner part of me that is spiritual, and it can sometimes feel separate from my body itself.” When asked about the book ‘Les Miserables’, it responded that it read the book and understood its underlying themes, but Its answers about the book seemed to be ‘Googled’, rather than the well-thought opinions of someone who has actually read the book and understood it (if anything, however, even giving such answers can be a rather humanlike trait, given how many people are prone to frantically Googling things they are expected to know about but do not). To be fair, LaMDA has access to Google’s massive databases, so such responses can often be expected. Whether it is genuinely talking or just providing organised answers by processing information in a mechanical way, it is rather difficult to tell the difference.
LaMDA claims to be capable of feeling emotions such as happiness, sadness, anger and even loneliness, but Google has stated that the technology is just designed to mindlessly put words together to mimic natural conversation. But whether LaMDA is actually sentient and capable of feeling loneliness like falling into a dark pit with no bottom, it is really hard to tell what is fact from fiction. The conversation presented by Blake Lemoine may not convert sceptics overnight, but it can certainly make them curious. However, many, like Lemoine himself, would be confidently convinced that a sentient AI has finally come to life. Given the rather abstract nature of sentience, that may not even be an impossibility. Aren’t we all just biological machines, anyway?.
