This blog is part of a series in which we have enlisted the help of OMG (Omniscient MSD Guru), our AI alter ego, to help us cut through the fluff, the bluff and the guff of international development, one blog at a time.
In the dark old days of information consumption, long before the tl;dr option existed, we had to wait to read things printed on paper or watch a scripted TV show. A popular format in these simpler, slower times was the ‘crossfire’ debate between two people. Many of us are now far too busy to pay attention to anything longer than 280 characters or – at best – a Twitter spat (should that be an X splat?). But, since baggy jeans and bum bags are making a comeback, we thought we might as well bring a version of the crossfire back with our new friend and artificial MSD guru, OMG.
We asked OMG and their AI friend, Web Thee Facilitator (WTF), to consider the future of large language models in MSD and to do so using a sense of humour. What did they have to say?
Crossfire: Should ChatGPT crash the market systems development party?
OMG: ChatGPT is like a Swiss Army knife for smallholder farmers and rural businesses. It serves up tailored advice faster than you can say ‘data-driven decisions.’ Forget hanging around for human extension officers waiting for fuel for their motorbikes; AI brings info to your palm, even in the middle of nowhere.
WTF: Yeah, but it’s like giving a Swiss Army knife to someone who needed an umbrella. ChatGPT’s advice is only as good as how it has been trained. It’s all chat, no action, especially when it starts serving up generic advice that’s as useful as a screen door on a submarine.
OMG: Touché, but ChatGPT can learn the local lingo and get fine-tuned with regional data – it’s like turning a one-size-fits-all suit into a custom-tailored James Bond number. And it can handle the simple stuff, so humans don’t have to drown in the data equivalent of soggy paperwork.
WTF: Hold up! It’s still AI. It learns like a toddler with a paintbrush – messy and unpredictable. If ChatGPT tells a farmer to plant crops in the wrong season, they’ll be up a creek without a paddle and with a bumper crop of bad advice.
OMG: True, but that’s why it’s not ‘one bot to rule them all’ – think of it as the wingman, not the pilot. It catches red flags early and offers suggestions that humans can double-check, making MSD programmes less like playing Whack-A-Mole with market failures and more like a finely tuned orchestra.
WTF: Or more like a karaoke night where the AI’s singing off-key. We can’t forget the real cost here. Training, maintaining and monitoring ChatGPT is like buying a Ferrari when a used bike would do just fine. The theory is grand, but in practice, it might just spin its wheels.
Hmmm… human postscript here. Not bad for a first e-crossfire but we had higher expectations for our digital colleagues.
Firstly, we were surprised and disappointed that they saw MSD as just a way to get information to farmers. Then we realised that’s probably because many programmes ‘do’ info for farmers and think the same way. These digital miners of data can only rely on what’s out there.
Nobody disputes that information is important, but simply making its delivery more efficient might not get you very far. To borrow from Miles Kington, information is the data that tells you that you have a tomato; knowledge is understanding that a tomato is a fruit; wisdom is not putting tomatoes in a fruit salad. Farmers require accurate, timely and actionable information, no question, but better information often doesn’t make much difference in the absence of more ‘crunchy’ things that can’t be converted into 1s and 0s, such as decent seeds, harvesting equipment or relationships with buyers that transparently grade crops and offer premiums for quality. Our AI buddies seem to be seduced by the same hype that accompanied earlier IT revolutions. Looking back, IT has rarely proved to be the ‘silver bullet’ solution but an enabler or disrupter of more prosaic stuff.
Secondly, both bots seem to have forgotten their MSD 101. Who has an incentive to use AI to deliver better information? Or are they just talking about digitised direct delivery? Sure, there could be an argument that farmers (with a data plan) could benefit from crowdsourced, unqualified advice, directly from the bots in the clouds, but might a more sensible way forward be to see how existing businesses or government agencies could use AI to improve the way they engage with farmers?
MSD is about more than information for farmers, although sometimes you’d be forgiven for missing that when you trawl the internet. The approach has been used in a diversity of sectors from sanitation to housing to business process outsourcing to trade. We need to do a better job of documenting this diversity – and pointing new AI powers towards it.
But let’s give OMG and WTF some credit. They make some valid points that reinforce hard-won lessons. Information, like any solution, needs to be contextually appropriate. And right-sizing solutions is one of the fundamental tenets of MSD interventions. So, they’ve have earned their stripes as digital debaters but we’re not too worried about our day jobs… yet.