Dialogue Neural Networks: How Chatbots Become Companions in Everyday Life

Since November 2022, when OpenAI allowed public access to ChatGPT, more and more new applications have been found for dialog neural networks. One of those is to mimic emotions and individual character in communication with people. A chatbot can conduct a conversation in the manner of Albert Einstein or Harry Potter, play the role of your best friend or a deceased relative. Artificial intelligence may even provide psychological support or help in completing a training course. In this article, we will explore a few examples of chatbots that elicit an emotional response and even affection, and talk about their pros and cons.

Neural networks are being used to support mental health. In 2021, 22% of the US adult population turned to their help — such figures were announced by experts from the Marquette University. Among the factors behind such interest, scientists note that neural networks respond 24/7, adjust to the user during a conversation, and a communication session with the network costs much less than a consultation with a specialist. People seeking psychological help use ordinary chatbots along with specialized applications such as Elomia and Mindspa.

Some development companies advertise AI-based applications as a full-fledged alternative to psychotherapy. And though these proclamations are usually accompanied by a warning of potential risks to the user, such practice raises concerns among experts. Researchers from Simon Fraser University highlight that neural network training is based on a limited amount of training data and cannot supersede the education and experience of a specialist. At the same time, the frequency of interactions and the illusion of communication with a human being evoke in the application users the trust and responsiveness even to the advice that may worsen their condition, the scientists note.

When neural networks are used in education by either teachers or students, the benefits go side by side with potential risks, according to a report from the World Economic Forum. The authors of the report indicate that AI in education holds immense potential: it can be used to create video and audio materials based on textbooks, adapt them for children with special developmental needs, or, for example, automate the checking of school homework. The discussed risks include the leakage of data received from students.

An example of the practical application of neural networks in education is Harvard AI Sandbox, a secure university environment for working with large language models. In the fall of 2023, Harvard associates received access to this environment for running their independent experiments. In their report with the experiment results, the University's Vice Chancellor for Educational Development, Bharart Anand, and his colleagues concluded that the neural network helped to improve the quality of training: with its help, the instructors generated interactive materials, created homework assignments, collected and analyzed responses of students in real time. According to the authors of the report, a major disadvantage consists in the so-called hallucinations, i.e. the errors of neural networks, as a result of which any of their answers require verification.

People expect the artificial intelligence to provide not only professional assistance, but also everyday human communication. Developments in this area are already underway: ChatGPT is able to generate on-demand responses corresponding to different social roles, and the virtual assistant Alice has a permanent personality, or rather can have one of several characters.

Some developers go further and offer applications that specialize in simulating close and even romantic relationships with the user, such as Anima, Kuki and Replika. In the latter, not only may the user chat with an AI companion, but also customize the appearance of its 3D model, and even purchase virtual clothes for it, including outfits.

Virtual companions attract criticism: for example, Sherry Turkle, Professor at the Massachusetts Institute of Technology, believes that they give the impression of a relationship, which may cause emotional disorders. Speaking of emotions, a researcher from the University of Sydney characterizes AI empathy as “cognitive,” and this type of empathy is characteristic of people having psychopathic disorders.

The neural network companions may act not only as presumably unknown interlocutors, but also may emulate certain people, for example, friends or relatives of a user. Some cases are known, when during funerals guests were offered to talk to a digital clone of the deceased: at the funeral of actor Edward Asner it was possible to say goodbye to his AI twin created by StoryFile. Matthew Asner, the actor's son, later commented on his experience, “He really was in front of me, the one I missed so much, my best friend.” However, Nigel Mulligan, a researcher at Dublin City University, holds a view that in exchange for temporary relief, such practices may disrupt the natural process of coping with grief.

Through special chatbots you can communicate with absolutely anyone: from your idols of childhood to the heroes of fiction. For example, on the Character.AI platform, a user can “meet” James Bond, Donald Duck or Sauron, not to mention hundreds of historical figures and present-day celebrities like Thomas Edison and Elon Musk. This opportunity is proving popular, as is apparent from 1.7 million downloads of the Character.AI mobile app in its first week of sales, and the company's revenue for 2024 of $16.7 million. In the app, not only can users interact with pre-configured characters, but also create their own virtual companion by configuring their personality and appearance.

Sometimes celebrities create their own digital twins to communicate with fans. For example, according to El País, that is what American singer Mark Tuan and boxer Francis Ngannou did with the assistance of specialized companies.

Virtual companions are the very latest evolution in chatbots for communication. They are popular with users, and the profits of the companies that offer access to them are growing year on year: researchers from Grand View Research predict that the market will grow at a rate of 47.1% per year between 2024 and 2030.

Despite the criticism, the digital avatar industry is progressing, and in the near future we will become witness to even more realistic and advanced solutions. Altego Digital Avatar is one such solution. By description of a visionary and a founder of Future Crew, Eugene Chereshnev, “It is a first-of-its-kind private artificial intelligence that learns from the digital footprint of its host.” The virtual twin will help to create content for your brand easily and effortlessly, and its ability to speak different languages will allow to embrace the global audience.