Home < Blog < The future of personal AI: Opportunities and Challenges in the Era of Personalized AIs

The future of personal AI: Opportunities and Challenges in the Era of Personalized AIs

April 17, 2024

By Hiroaki Hamada, Senior Researcher in Araya Inc.

Science fiction paints tantalizing visions of futures filled with advanced AI. Today, with the emergence of Large Language Models (LLMs) and their capabilities, we stand on the precipice of a new age where personalized AIs are not mere aids but vital extensions of our society. This series will delve into emerging technologies that potentially support human Well-Being and the background underpinning them. In doing so, we will dissect emerging technologies from different perspectives encompassing AI, Web3, Well-Being, and ethical considerations. In the first article, we will spotlight the opportunities and challenges tied to personalized AIs, focusing on the Second Brain and AI assistants.

Co-existence of AI and Humans are Already Coming

Have you heard of the novel, “Klara and the Sun” (2021), written by a Nobel laureate, Kazuo Ishiguro? It is a story about the temporal friendship between a little girl, Josie, and her personal AI robot, Klara. Klara is not an up-to-date AI agent, but she is designed to observe and support children's development and mental health. In the end, Klara’s observational qualities turn out to be critical. Ishiguro depicted an upcoming world in which humans and AI agents co-work. Ishiguro depicted a vision of a near future where AI plays a significant role. The vision resonates with the advent of language AIs such as Large Language Models (LLMs).

LLMs and Personal AIs

LLMs, such as Bard and GPT4, are a type of natural language processing model, which require vast amounts of text data mainly sourced from the Internet for its training. LLMs can perform multiple linguistic tasks such as translation, text revision, and text summarization, and have demonstrated their utility in numerous applications. A study suggested that LLMs improved writing efficiency and performance inequality [Noy&Zhang, 2023]. Although the training of LLMs demands thousands of GPUs and power consumption, their deployment can be achieved on home computers with GPUs. A further strength of LLMs is the capability to fine-tune their output with smaller datasets including laboratory data and personal data. This fine-tuning unlocks the potential to leverage datasets with high data privacy requirements. Two types of personal AI, AI assistant and Second Brain, are under exploration to uncover new use cases. In this article, we will dive into the potential that these personal AIs may offer.

What are AI Assistants?

AI assistants have become an integral part of our daily lives, offering services that support a variety of tasks such as information retrieval, slide creation, and scheduling. These digital helpers use sophisticated technology to simplify our tasks and make our lives more organized. Examples of AI assistants that many of us are already familiar with include Siri from Apple and Alexa from Amazon. These tools utilize Natural Language Processing (NLP) to understand and respond to our verbal commands, thereby facilitating smoother and more natural interactions. When LLMs are incorporated into these AI assistants, they can deliver even higher comprehension of natural language. LLM-backed AI assistants are trained on an extensive range of language data, which allows them to understand and generate text more effectively. Moreover, they can leverage user-specific information to personalize responses, making the conversation feel more optimized and engaging. This enhancement in natural language understanding not only improves the user experience but also extends the range of tasks that AI assistants can handle, further elevating their utility in our daily lives.

Second Brain as an Extension of Individuals

Second Brain, originally from Tiago Forte’s book, “Building a Second Brain, " is a concept primarily used in a personal management context. The idea of the Second Brain is about one’s digital twin to visualize, reorganize, and make new connections with things one learns information and generates thoughts. Offloading information and ideas externally allows us to efficiently and effectively manipulate information and enhance creativity. Note-taking tools such as Notion, Obsidian, and Evernote are often used to create Second Brain. The advent of LLMs provides a new possibility for Second Brain in which information storage and manipulation is automated. For example, REWIND AI, a software company, provides a tool to keep all screen information of a PC and mobile phones and to search information through personal LLMs. Essentially, this LLM-backed Second Brain automates informational storage without one’s note-taking. Whether this service itself is sustainable or not, but opens up an opportunity to record screen information and manipulate it efficiently. This LLM-backed Second Brain can be further used for collective manipulation to promote generating ideas.

Difference between AI assistants and Second Brain

Although AI assistants and Second Brain share a common ground of manipulating information and generating ideas, they may differ in their information generation approach. The AI assistant can predict outcomes or recommend actions from the perspective of different experts while Second Brain focuses on interpretation and actions from more subjective perspectives. Hence, the two systems are complementary since they enable us to support idea generation from objective and personal perspectives.

Issues on Personal AIs

Potential issues such as data privacy and response appropriateness are noteworthy. As AI systems optimize their response based on personal information, there are concerns about it when companies extract personal information and utilize other purposes like advertisements. Additionally, the potential that LLMs yield any responses including true and false information with justification is problematic. The problem is known as “hallucination”. Therefore, higher data privacy and an appropriate response filter are necessary.

Potentials of Personal AIs for Well-Being?

We are now in a world of information overload in which we cannot digest all information without abstraction and focus. In the era of language AIs, use cases of assistants and Second Brain with LLMs or similar kinds will be explored to support our idea generation. Mental health is one of the main targets of LLMs due to the human nature of sociality. In the following article, we will dive into the potential of AI to support human Well-Being. In doing so, we will grasp the social roles of AI and its required technological and legal infrastructure.

References

Ishiguro, K. (2021). Klara and the Sun: A novel. Knopf.
Noy, S. and Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science. 381 (6654). pp. 187-92. doi:10.1126/science.adh2586
Forte, T. (2022). Building a Second Brain: A Proven Method to Organise Your Digital Life and Unlock Your Creative Potential. Profile Book.
REWIND AI. https://www.rewind.ai/

‍