#artificialintelligence #80: Setting up an AI Research Lab: Values and Principles

This week, I am in India and we will announce more about our work on AI in Agtech?

Leveraging our work at the University?of Oxford AI in agtech course,?and also in collaboration with our partners ADC Baramati and Microsoft farmvibes, we are setting up a grassroots initiative for AI and agtech based on the recently open sourced Project farmvibes from Microsoft

I have discussed the farmvibes initiative before #Artificialintelligence 77 farmvibes data driven agriculture

The initiative will be set up by some of our team in Oxford working with teams from Microsoft and ADC Baramati

The idea is to set up a local center for agtech and to create demos in farmvibes based on some compelling use cases around themes such as microclimate change, causal analysis in agriculture, satellite data in agriculture etc

Hence, my interest in understanding how to set up an AI research?lab

Ultimately, such an initiative is a grassroots institution.?

And any institution should be a self sustaining ecosystem - which should span beyond the individuals?who set it up

So, the starting point of such an initiative is the values and principles

Here is my list for setting up a grassroots initiative for an AI lab

Welcome thoughts

1)?Narrow the domain / problem statement specifically within AI with a yearly focus (ex specific?algorithms, data types etc) - ex: 'what if/causal' in our case for AI and agtech

2)?Physical spaces and requirements which can be tied to the AI ex hydroponics, climate?change sensors?etc

3)?Datasets (new and existing) including?creating synthetic datasets

4)?Emphasis on rapid experimentation - small changes and small experiments which can show results and that can be built upon

5)?A feedback mechanism from industry / AI domain experts

6 ) An education program for developers and domain experts working together on a specific problem

7)?Emphasis on rapid publishing on github (instead of traditional academic publishing - a process which is arcane and broken)?

8)?Addressing large and systemic?challenges for AI ex microclimate prediction

9) Mentorship

10) Sharing best practices?globally?

11) Incorporating expert/domain?knowledge

12)?Ability to read and synthesise research papers including the ability to understand literature reviews and how to build?upon a body of knowledge?

13) Understanding the state of the art in any domain and how that can be improved in AI

14) Statistical and maths knowledge (especially when there is less data) and an understanding of when to use deep learning techniques (ideally suited for highly parameterised models with unstructured?data)

15)?Explainable and Responsible AI

16)?autonomous systems?

However, the biggest challenge is in developing and nurturing grassroots?talent with a new mindset

Today, I attended an event organised?by the Ashoka?University - where a speaker said that

'jack of all master of none' is a false dichotomy. what we need is a belief where people need to be masters of lifelong learning so that they can master multiple skills rapidly.

Inculcating that mindset for AI is the biggest challenge for setting any AI labs as a grassroots?initiative

Hence, the emphasis on values and principles

welcome your thoughts

If you want to study with me at our courses at #universityofoxford for #artificialintelligence, #digitaltwins please see https://lnkd.in/eSC2scZ and https://lnkd.in/dRewYpPC).?

Image source: upsplash??https://unsplash.com/photos/yovhXPl8V1M