DailyPost 2947
FROM CHATBOT TO AGENT BASED SYSTEMS TO AGI
Whether we understand AI or not, one thing is for sure that we all understand that it is transformative way beyond what we can imagine. Post ChatGPT, the utility of Artificial Intelligence has been established and all of us are being impacted in some way or the other. The general feeling is that it is all for our good and it would take drudgery and time taking mechanical / repetitive work out of our lives. We also accept the fact that the final destination is to achieve artificial general intelligence or AGI – “a still hypothetical form of robot intelligence that mimics human abilities.”
What is the road to reach there is the issue. There is nobody better than Dennis Hassabis, founder DeepMind, creator of AlphaGo to AlphaChip and recipient of the 2024 Nobel Prize in Chemistry to provide the contours and the road ahead. According to him there is no secret formula to reach there. What is our present stage? We might not be able to pin it down exactly, but we all know for common usage we are at the chatbot stage, best manifested by most common ChatGPT. This is the conversational form of the GPT, which itself is improving exponentially with every version.
What does the ChatGPT and other similar systems (LLMS) do? Such systems passively answer questions, simultaneously having the capability to summarise text and also does instantaneous research unheard so far. The next stage would be the agent based systems. What traits and skills would they have? They would be in a position to plan for example they can think ahead, plan a trip or book tickets. It should be competent enough to take action in the real world. Reasoning skill of AI agents would be a game changer. They would be able to reason through problems. AlphaGo is an example. What the next phase of agent-based systems would apply these skills to real-world problems.
The above skills sound quite a leap from the Chatbots. The agent-based systems should possess better memory, they need to remember the details they have been told. Outstanding personalisation would be the name of the game. They may literally merge in you, like understanding the user’s preferences, likes and dislikes. They will be able to use hardware like robots in the physical world, any type of software and even other AI systems. Even after using LLMs, these traits and skills sound sci-fi. It is in the offing for sure. On the issue of the timelines to reach this and beyond to the final destination of AGI, Demis Hassabis thinks the full realisation of an artificial general intelligence that can reason as well as the humans is still a decade away.
IF AGI WERE TO HAPPEN IN A DECADE, EVERY MILESTONE IN THAT JOURNEY WILL ELEVATE HUMAN EXISTENCE TO A DIFFERENT LEVEL, AS CHATBOT HAS!
Sanjay Sahay
Have a nice evening.