The Journey to Generalizable Intelligence – The Top 5 Takeaways3 min readReading Time: 3 minutes
July 12th, 2022 saw multiple AI experts and industry leaders gather at Multiplai, Target India’s conference on building future-ready retail with the scaled application of AI. Anand Chandrasekaran, the Founding Chief Technology Officer of Blox AI, took the stage to deliver the talk “The Journey to Generalisable Intelligence” breaking down ‘general’ and ‘artificial’ intelligence and the future of AI-powered platforms in the world of retail.
Anand elucidated his thesis about the future of intelligence platforms, or rather, the journey towards it, peppering in analogies from the world of sports, retail and even wildlife! Here is a quote-by-quote breakdown of the 5 takeaways from the talk that shouldn’t be missed!
Takeaway 1: “Intelligence is a made-up word and we have tried to force it”
What is intelligence? Broadly defined as the ability to understand, learn, think and process sensory information to act upon it. Using the example of deer in the wild, Anand broke down intelligence and touched upon the features of intelligence that aren’t discussed too often. And these are:
These are the features that set human intelligence apart and enable meaningful decision-making across different scenarios.
Takeaway 2: Artificial vs General Intelligence – or neither?
“Artificial, General as prefixes should be thrown out, our focus should be on intelligence,” says Anand. Models being built today by data scientists across the world are getting closer and closer to neural architectures, and the intelligence that is being built will be very human – not just in the way the algorithms work, but also in how it responds to stimuli. The goal is to get machines to process signals and carry out tasks the same way a human would to enable faster and reliably cheaper operations.
Takeaway 3: How do we learn?
Humans learn in different ways. What are they and why does it matter for intelligence platforms?
- Serendipitous learning – Learning that occurs through a progression of events that is reinforced depending on the outcome being favourable or unfavourable. For example – The way humans learn to walk.
- Mimicking – Learning that occurs as a product of our environment, where we mimic and emulate our parents and peers.
- Instruction – Learning that occurs when we are given information and or directions
- Repetition – Repeating actions until you’ve learnt it well enough and your brain wires itself to perform the action with ease.
- Experience – Experiential Learning occurs as you collect and process information as you perform tasks or actions. “Learn by doing”, to put it in colloquial terms.
Now, why does this matter in the context of intelligence platforms? That brings us to the next takeaway – an important one.
Takeaway 4: A Singular platform
The ideal intelligence platform is singular – one platform that can solve many problems. This platform will possess all the types of learning discussed above to simulate what humans are doing.
This platform can collect and process data and information, store it, learn from it, reinforce it and adapt to different scenarios and use-cases based on context. The platform will accommodate different types of models to solve different types of problems and application layers through which it will interact with the world, enabling the solving of problems across domains – be it supply chain or personalization.
Takeaway 5: The principles to build a singular platform
The principles to build a singular, intelligent platform is to bring all the three necessary layers under one roof – like building a brain. And the three layers are:
- Infrastructure layer
- Model Development Layer
- Application Layer
The infrastructure supports the flexibility that the model developers have and the business use case wins all.
With the advent of AI becoming ubiquitous and leaving a mark on our day-to-day life, an all encompassing intelligence platform could help businesses across domains take major strides in making the world a better place.