Opinion
The evolution of AI: The balance between companionship and efficiency
Summary
- Given today's multifaceted interactions, a diverse AI ecosystem seems imminent, each varying in companionship and efficiency levels
Throughout our lives, we continuously interact with various individuals and organizations – from family and friends and organisations to global corporations like Google, Microsoft, and Netflix. While we form opinions of these entities, they too shape views about us. Oftentimes, we hope for more understanding, even sometimes compromising our privacy for more efficiency in services – be it Netflix's series recommendations or Amazon's targeted product suggestions.
Traditionally, both humans and corporations have leveraged their own intelligence in these interactions. However, the advent of Large Language Models (LLMs) has intensified the availability of intelligence. This artificial intelligence (AI) can be classified into three categories: Peer-to-Peer AI (P2P AI), which aids in personal tasks, such as drafting emails or messaging; B2C AI which enables direct consumer interactions, like chatbots assisting in booking tickets; and B2B AI where the system focuses on data analysis, predicting energy consumption or managing logistics.
For P2P AI, the "Companion Style" approach is crucial, which really understands the human beings who are engaging with it. Every individual would like a companion-styled AI which can help the individual achieve certain goals- like writing a personalised note, understanding an academic concept or even adhering to a fitness goal. For such an AI to be effective, it needs to understand the unique traits of the individual, her likes, her preferences and knowing how to nudge and help the person to achieve a certain goal.
A B2B AI primarily deals with data patterns and numerical analysis to make informed decisions. Such an AI would need to prioritise efficiency over emotional considerations, largely working behind the scenes to achieve its goal.
A B2C AI system is where things become tricky and one can argue whether it should appear like a companion or an efficient robot. As much as we do not want to have a conversation with a robotic entity, given that an AI is trained on a vast amount of data – including novels, dialogues, movies, books on psychology etc. – it might understand how to hack the emotions of a human being and it can lead them to actions that may not be in their long-term interest like maybe spending indiscriminately on unwanted items or nudging them towards a dangerous doctrine.
As AI today surpasses Turing Tests – designed to differentiate between machines and humans – the risk of emotional manipulation by AI increases. However, an emotionless, hyper-efficient B2C AI might be as effective as a mere warning sign in isolation.
Given today's multifaceted interactions, a diverse AI ecosystem seems imminent, each varying in companionship and efficiency levels. The ideal scenario? A holistic AI, much like Jeeves, embodying the best of understanding and effectiveness.
Pawan Prabhat and Paramdeep Singh are co-founders of Shorthills.AI.
