The rise of AI in the healthcare sector is notable, with the industry surpassing $40 billion in the U.S. in 2022 and projections suggesting it could reach $1.3 trillion in the next ten years. A recent report anticipates that the market for mental health and therapy chatbots will grow to $1.71 billion by 2031. These tools provide essential support and guidance for various mental health challenges, including grief. With mental health issues affecting approximately one in four adults and one in ten children annually, established companies in this field include Woebot Health, Wysa, Minstrong Health, and Marigold Health.

Grief technology is transforming our interactions with those we've lost and how we navigate the grieving process. Our digital footprints—text messages, photographs, videos, emails, and social media interactions—help us maintain connections with loved ones, regardless of distance. AI voice cloning technology enables the creation of digital versions of human voices by analyzing distinct speech characteristics such as accent, tone, and pitch.

One significant advancement in using AI in therapeutic settings is the development of language model chatbots. A prime example is ChatGPT, created by OpenAI, which serves various human needs, including generating outlines, collaborating on content, and providing mental health support.

According to a 2021 survey by Woebot Health, 22% of adults reported using a mental health chatbot. Economic barriers often hinder access to mental health care, and AI chatbots may lessen reliance on human therapists, potentially widening the wealth and health gap for underprivileged communities. The rapid pace of AI advancements often outstrips research capabilities, resulting in limited sample sizes that can lead to biases in AI data.

Tess, a mental health chatbot, is marketed as a therapeutic or coaching tool. It aims to uplift users, alleviate anxiety, and provide various resources to promote mental well-being. Utilizing emotional algorithms and machine learning, Tess interacts with users through adaptive technology. As one of many "robot therapists," Tess is employed alongside traditional therapy to alleviate the demands on overstretched health services. The advantages of mental health chatbots include their cost-effectiveness and accessibility for those hesitant to disclose personal issues to a stranger. Unlike conventional cognitive behavioral therapy (CBT), chatbots are available 24/7 and offer a judgment-free environment.

In the realm of mental health, natural language processing has proven effective in identifying and categorizing various mental health conditions such as grief, depression, and anxiety. The COVID-19 pandemic has intensified the demand for mental health resources, positioning chatbots as a vital support tool.

AI technologies like ChatGPT and other large language models could potentially simulate conversations with deceased individuals, introducing complexities in estate planning. When a person passes away, their possessions are distributed among surviving family members or donated. Five years after my mother's death, I still navigate my parents' belongings, slowly digitizing photographs. Each encounter with their memories can trigger grief, prompting me to postpone the task yet again.

Death technology, including AI-generated videos or bots created from remnants of our loved ones, necessitates the involvement of either the deceased or their survivors, often requiring a subscription to preserve these interactive memories. Many such companies may eventually fail or become outdated, leaving users to transfer these digital memories to future technology.

Hollywood and Silicon Valley often pioneer new technologies, with recent AI death tech companies emerging that promise to recreate the "essence" of our departed loved ones. Startups like StoryFile, Replika, and HereAfter AI offer various services to assist those grieving. For instance, StoryFile enables interactive video conversations with the deceased, while Replika serves as a companion chatbot that evolves with user interaction.

This burgeoning niche of AI death technology employs deep learning and large language models to replicate the likeness, speech, and personality of the deceased. Conversational AI facilitates communication between machines and humans, while generative AI utilizes machine learning algorithms to create new content, including videos, animated images, speech, and text.

Grief tech platforms often implement tiered subscription models, which can be costly. My free subscription to StoryFile resembled outdated video formats, starkly contrasting with the high-quality promotions on their website. For a one-time fee of $499 annually, users can access enhanced video features of their loved ones.

Seance AI allows users to engage in online conversations with the deceased, with founder Jarren Rocks describing the service as a means for the bereaved to achieve closure, rather than a tool for maintaining long-term connections. The concept behind You, Only Virtual is to never bid farewell to our loved ones. Founder Justin Harrison proposes the audacious idea of completely eliminating grief, while simultaneously claiming to reproduce "the authentic essence" of those who have passed. However, if essence is tied to our identity, then eradicating grief complicates our humanity.

China's AI sector is also making strides, with several companies developing griefbots. Super Brain, tailored for the bereaved, offers visual and audio clips along with conversational video chatbots that imitate the deceased. These customized griefbots come at a steep price, ranging from $6,860 to $13,710.

Nanjing Silicon Intelligence (NSI), leveraging Huawei’s Pangu language model, seeks to merge digital avatars with large language model datasets to achieve digital immortality for entertainment purposes. NSI co-founder Sun Kai developed a video-enabled chatbot platform designed for individuals to communicate with lost loved ones. In China, regulations governing AI require consent from individuals whose personal data is used in deepfake creation, though the specifics of enforcement remain unclear.

To address these concerns, UNICEF's East Asia and Pacific Regional Office has created a guide for safer chatbot implementation in a region witnessing a surge in chatbot adoption as substitutes for mental health practitioners. North America currently leads the mental health chatbot market, with initiatives aimed at addressing gaps in mental health care.

Futurist Ray Kurzweil is at the forefront of griefbot technology, predicting that individuals will connect their brains to machines—a concept known as singularity—by 2045. After losing his father at 22, he is attempting to replicate him by feeding an AI system with his father's writings and compositions. More ambitious plans involve using nanotechnology and DNA from his father's remains, though this requires considerable effort. Kurzweil's "Dad Bot" would engage him in discussions about his work. From personal experience, I find this approach dangerously close to fostering prolonged grief. If we remain entrenched in the past lives of our departed loved ones, how can we fully live and embrace our own lives?

Amit Roy-Chowdhury, a computer engineering professor at UC Riverside, expresses skepticism about our ability to create AI replicas of the deceased that can convincingly engage the living.

> "All artificial intelligence relies on algorithms that must be trained on extensive datasets. If ample text or voice recordings of a person exist, it is feasible to create a chatbot that mimics that individual. Challenges arise in unpredictable situations where the program encounters new scenarios. Many AI systems merely memorize patterns and lack a deeper understanding that would enable them to generate novel, yet appropriate, responses."

> "When observing advanced AI applications designed to replicate real individuals, such as in the documentary about Anthony Bourdain, we often misjudge AI capabilities. That project succeeded due to the vast amount of Bourdain's recorded material in diverse contexts. If data is available, it can train an AI to behave within learned parameters, but it struggles with unique or irregular events. Humans possess a broader semantic understanding, allowing for entirely new responses. The complexities of semantic processing are substantial."

> "In the future, we may develop AI that responds to new situations like a human, but the timeline for this advancement remains uncertain. Debates within the AI community reveal differing opinions on whether this will take over 50 years or if we are closer than anticipated."

Ongoing legal and ethical dilemmas surrounding grief technology pertain to consent, data privacy, psychological dependency, biases in datasets, economic accessibility, and the risk of technological obsolescence.

New York has enacted regulations governing post-mortem publicity rights, primarily for celebrities. Technology advocates are suggesting a "Do Not Bot Me" clause in estate planning. Some states, including California and Connecticut, have introduced data protection laws to safeguard user data collected by corporations. The European Union's General Data Protection Regulation imposes further requirements that could impact grief tech services.

It is crucial to recognize that commercial grief tech companies operate as businesses, potentially misleading consumers into believing they offer personal support during traumatic experiences. While not all grief tech companies are exploitative, they are businesses responding to consumer demand. Similar to photography, which helps us maintain connections with the deceased, the evolving technology in this space is still under investigation regarding its effects on those in mourning.

Ginger Liu is the founder of Ginger Media & Entertainment, a Ph.D. researcher focused on AI and visual arts media, particularly grief tech, digital afterlife, AI, and mourning practices. She is also an author, artist, photographer, and filmmaker. Listen to her podcast on The Digital Afterlife of Grief.