AI and public policy

One of the things I signed up for, is a course on AI and public policy. After 20 years of data and analytics in the marketing sphere with the last year and a bit of dabbling in AI or rather the data and analysis behind making marketing AI driven, it is but natural to wonder why the sudden love for public policy?

Truth be told, I do not have a clear answer. I just felt it was something that I had to do. Or maybe deep down, I truly believe that one of the pitfalls of AI is going to be increasing the divide between the haves and the have nots. And I definitely, am not solving for those pitfalls but a little knowledge doesn't hurt.

One of interesting things about an education in this day and age is that people assume you know and will use LLMs whether it is to craft responses, quickly look up information you did not know about. I used it for 2 things, the first to ground me in my journey through the public policy course. The second to generate an image for this blog.

While the second took literally a second (apologies could not resist that one), the first was interesting. I referenced the fact that everyone assumes you will have access and use an LLM during the course. The course coordinators went a step further and helped us craft a very engaging prompt to which we could upload all our course material. Engaging since one of the instructions as part of the very elaborate prompt was "guide me to discover insights rather than just giving me the answers". This makes the journey far more interesting with tidbits and nudges rather than direct staid answers.

ANyway, the more I dwell on this aspect of public policy, the more I fear, my original thoughts about AI increasing the digital divide in the world is not some product driven hallucination. You see as used in business, the focus of AI is efficiency, improve processes, increase sales/profits/revenue, reduce costs, shorten workflows whatever not but when it comes to policy decisions, the focus is not efficiency but more ensuring that all sections are being covered - that is equity. So this becomes an efficiency vs equity problem. The AI model will always try and go for the more efficient way to solve for a problem, the policy focus is to ensure a slower but more deliberate process which encompasses any 'dis-efficiencies' (is that even a word?) within.

The second aspect is data, something I am more familiar with. Any AI driven model needs to train on data, the more data it trains on the better it becomes. Unfortunately, when it comes to public policy, the least fortunate amongst us would be generating the least data and thus the policy which should be to help bring them up the curve would be unable to do so as there is very little data about them. And thus the second conundrum efficiency vs legitimacy.

I don't know how to solve for these, after all it has just been week 1 of the course, but I leave you with a very interesting example which was discussed during one of the sessions. India has a vast coastline and with such a long coastline, comes a plethora of fisherfolks. Now a lot of these fishing tribes are those who have done this as a profession generation after generation. A lot live in poverty, not even able to afford motor boats forget trawlers. Yet they continue to ply their trade because they know not anything better. Now that I have set the context, the ideal policy to get them out of this low skilled labour intensive profession is to get them loans to buy better boats! But here is what actually happened. India as a country had a very poor landline penetration. When mobiles came along though, the focus was not on increasing landline penetration anymore and this combined with cheap data plans and being the amongst the lowest global cellular tariffs ensured that mobiles permeated far and wide, into each village and now the fishing tribes had access to data which allowed them to understand which markets around them had higher prices and thus they could thus divert their meagre stock accordingly. It did not bring them riches but did make them owners of their destiny in a tiny little way.

I end here, will write more as the course progresses but what strikes me is not that we should be getting the highest number of GPUs or building the best AI talent or starting the most valuable AI companies but ensuring that one can use AI to promote equity and information diffusion can be one of the biggest ways to do that.