Mustafa Suleiman, Microsoft AI CEO, said that models expected to seem in 2025 may have almost infinite memory. Through this, the interaction between humans and artificial intelligence (AI) will fundamentally change, and it is predicted that AI will eventually develop into a ‘life partner’.
CEO Suleiman recently appeared in an interview with Times Tech, a technology media outlet, and said, “We’ve got been working on a prototype with almost infinite memory.”
Infinite memory refers back to the ‘long-term memory ability’ of AI chatbots. Following OpenAI’s ‘ChatGPT’ which introduced this, Google also added it to ‘Gemini’ from the nineteenth.
MS plans to expand this almost infinitely. Through this, we consider that AI can develop into a life partner that plays the identical role as a teacher or personal doctor based on conversations with individuals throughout their lives. He explained, “That’s why I’m a co-pilot.”
He also said the feature “means it’s value it for people to speculate time talking to chatbots.” Because “the whole lot you say will come back to you in a useful way in the long run.”
Ultimately, chatbots will provide support and advice to users, helping them with the whole lot from planning their day to deciding methods to live their lives.
In fact, it should cost money to expand memory at first, but because the variety of users increases, the fee will decrease. In reality, it was revealed that the fee of using the Microsoft Azure model has develop into 99% cheaper in only two years. He also explained that with the advancement of algorithms, more things have develop into possible.
This was also included within the ‘Suggestions for using ChatGPT well’ revealed by OpenAI product manager Nick Turley the day past. He said that should you don’t use ChatGPT just for asking questions but tell people what you need to know, it should be effective later.
CEO Suleiman expanded this to the unit of life. Personalized chatbots can remember previous conversations and understand the context of life even many years later. Based on this, you too can predict what you would like.
Particularly, before joining MS, he founded Inflection AI and developed the emotional chatbot ‘Pi’. Unlike other chatbots that only provide accurate answers to questions, Pi has the flexibility to ask in regards to the user’s mood and proceed the conversation in line with the user’s tone of voice.
Also, as revealed on today, MS could be seen as developing its own model individually from Open AI technology.
When CEO Suleiman was hired last March, it was reported that Microsoft would develop an interactive model called ‘MAI-1’. This model has not been made public yet, however it has been confirmed that they’re currently working on long-term memory functions.
As well as, Microsoft announced the day before that it had signed a licensing agreement for non-fiction books with global publisher HarperCollins. It’s unusual for MS to sign a copyright agreement with a person publisher, which appears to be an try and learn historical facts from My-1. It’s because with a purpose to develop into a life partner, you should know what actually happened.
Within the interview that day, CEO Suleiman said that he was also impressed by the innovation of voice models that create a smooth and natural conversation feeling like Google’s ‘Notebook LM’.
Summarizing this, it could be assumed that the chatbot that MS will release next yr shall be a customized emotional chatbot that not only remembers all chats with users, but in addition knows what happened on the time, and communicates with users in a practical voice.
Reporter Lim Da-jun ydj@aitimes.com
