Nice article with some good suggestions, but I disagree with the majority of the suggestions given. 😀
An elitist strategy for AI, building centres of excellence and so on, will not give any competitive edge to India. It will simply prepare poaching grounds for US companies.
An elitist strategy for AI, building centres of excellence and so on, will not give any competitive edge to India. It will simply prepare poaching grounds for US companies.
The fact is there are already a few centres of excellence in India, and they regularly publish at top AI conferences. They are also doing the hub and spokes model for disseminating AI know-how to industry. The quantity of impact is small, but proportional to the investment made.
There is a lot of buzz-wording in the article, as AI has indeed become a hype in the industry. But the reason why India must invest heavily in AI and develop know-how is not because of industry.
It is to protect our democracy. Without a democratic AI, there will be no democracy.
It is to protect our democracy. Without a democratic AI, there will be no democracy.
Towards building a democratic AI, an elitist strategy is exactly the opposite of what must be done. So why are the AI giants like USA or China not doing it?
Because the initial trends in AI development were from military research, and this culture continues even to this day.
Because the initial trends in AI development were from military research, and this culture continues even to this day.
As a sovereign state, India needs to invest in AI for military objectives. But importantly, we cannot afford to build an AI culture in the image of the military. India (and EU, if it can get its act together) will bring important balance to AI culture if it can think differently.
So how would a democratic AI culture look like?
1) First and foremost, it will promote data sovereignty and data ownership for the citizens of the republic. The citizens cannot be data slaves to a military organization, or to a private (or worse, foreign) company.
1) First and foremost, it will promote data sovereignty and data ownership for the citizens of the republic. The citizens cannot be data slaves to a military organization, or to a private (or worse, foreign) company.
2) There has to be long term investment for sustainable AI. What is “sustainable AI”? AI models, hardware and algorithms that have low to moderate power requirements and data requirements. We should promote organic growth of applications around such models.
3) There must be a direct integration of AI with research on human cognitive development. The objective for AI algorithms cannot be just quantifiable evaluations on (dead) datasets. Instead, we need live evaluations, similar to games with humans in the loop.
4) The AI models need to be integrated tightly with interactive human computer interfaces. This is a natural follow-up from the previous objective. Certain applications in accessibility, elderly care, education, factory work etc. will naturally lead to such AI interfaces.
5) There must be an AI commons, both data and algorithms, that must be defined and defended. This AI commons needs software licensing and other legal instruments. Currently, the legal framework is extremely out of date, even in countries like USA, not to speak of India.
6) Ultimately, the spoken languages should be integrated within the AI models. This means not just NLP and speech recognition, but a concrete improvement of the human language itself! Similar to Sanskrit poetry, compound words and grammatical forms must arise for AI commands.
7) All of this needs to be achieved democratically, and cannot be left to a few elitist institutions. In fact, an elitist culture will never produce such a transformation in our society and culture. So this AI research and education must be broad-based, starting from schools.
8) This brings me to the most important point: a democratic AI education can only be in the languages of the people, not English. All the centres of excellence and top universities must offer AI courses in Indian languages. They must publish research findings in Indian languages.
If we don’t care for democratic AI, but just want a share in the global pie of AI economy, the research excellence is completely secondary. The strength in AI comes exclusively from data monopoly. The strength of US and Chinese companies is based entirely on the data they hoard.
So building academic centres of AI excellence is completely useless if these centres don’t have a greater access to the data (what I termed as “AI commons” in the above) than what is available in private companies. Researchers will simply migrate to where the data is available.
This is not a hypothetical thing that might hypothetically happen to an Indian centre for AI excellence. It is already happening in all too universities and institutions in US or Europe. Researchers are poached, not only through fat salaries, but through access to data.
It is just a waste of time to build such elitist institutions for AI, if there is no plan for data aggregation and management in the academic setting.
I want India to focus on the greater objectives of democratic AI. Everything else follows from there.
I want India to focus on the greater objectives of democratic AI. Everything else follows from there.
I gave the same comments on European initiatives towards building AI centres of excellence. An elitist strategy is completely useless if that is not coupled with an organic development of an AI commons (both data and software).
