I'm getting worried about AI
Categories: AI
Dario Amodhi has been in Australia scare mongering again. It’s not cool and his sentiment is leading to growing hostility towards the technology and the industries leaders. With sentiment like this, it’s no surprise that survey data is also showing younger people are becoming more anxious and less optimistic about what AI means for their future and a growing discomfort with how this technology is being pushed and what people believe it will do to them.
The message from the tech industry is AI is coming. It will reshape work. It will change the structure of the economy. You need to get on board or fall by the wayside. Alongside that is an assumption or an arrogance that the people building these systems understand what is coming in a way that others do not.
But it seems that most people are not engaging and the ones who are are not using some advanced, well integrated AI system. They are on the whole, experiencing AI using the free version of ChatGPT at home or Copilot at work because it has been switched on for them by IT. Understandably, these tools are what is shaping the broader AI opinion for many. These are not good experiences. They are low quality and frequently wrong and they exasperate the gap between what is promised and what is delivered. They give a glimmer of what might be possible but they're not the nirvana they're sold as. But people are being told to reorganise their work around this comming capability. Consultants are turning up and telling businesses to get their data AI ready (moving on from analytic ready). Industry organisations are setting up entire programmes, again scoped around data migration, restructuring, and governance so that AI systems can be layered on top. This is happening in agriculture as much as anywhere else. The pitch is the same as its been for the last 20 years in precision ag. Clean up your data, change your systems, invest now or fall behind. But isn't this what AI is supposed to deliver, isn't it supposed to fix all this stuff? Apparently it is, but it isn't. At least not yet.
While I’m definitely on board the AI train, I worry that It looks like a lot of effort being asked of people before the value has been proven. Lots of FOMO without the proven case studies and importantly, lots of people, mostly in America who are making a lot of money from the tokens we burn trying to figure out how to make it fit, how to make it useful in our businesses. We’ve been here before. Especially in agriculture and I’m worried we’re going to screw things up again and put a lot of people off what could be a very valuable, transformative opportunity. Especially so as I see new players coming in, going through the same loops again.
As with previous runs around the loop, people are being told AI will replace parts of their work, while also being told they need to use it to stay relevant. They are asked to invest time and effort into preparing for something that, seems complicated, time consuming for what to date shows a glimmer of payoff but only in fairly specific tasks like search, writing, summarising etc. It is not hard to see how this along with all the noise created by those hyping the next big thing has the potential to lead to apathy. I've told a few people that the situation reminds me of a scene from the Incredibles where the bad guy, Syndrome tells Mr Incredible - "When everyone has the power of the Supers, nobody will be!". This could be more true than I thought, especially as worries about cognitive decline and other psychological side effects of AI use we yet no little about potentially start to kick in or get pushback.
The data is also starting to reflect this. Among Gen Z I read, excitement and hope around AI are declining. Anxiety and anger are increasing. People are starting to feel trapped between two outcomes that are both pretty unappealing (get on board or get left behind). This is not how technology adoption happens and when I reflect on how agriculture approached technologies like big data I can start to see where we went wrong. It sounds really obvious but, when something works, people adopt it because it makes their job easier or improves their outcomes. The change follows the benefit. What I see happening here is, like the data discussion is the reverse. The expectation of change and opportunity is arriving first, backed by a narrative of inevitability, with the benefit still being argued for. I don’t think that more information and more software is what the industry needs. If anything it's a lot less. But at a much higher quality.
Farmers are already managing a complex set of responsibilities both on farm and in the office. Production, finance, labour, compliance, weather, markets. Adding more technology that requires them to learn new systems, restructure their data, and rethink how they operate is not a small ask or one they will particularly be excited to do... again. It is another demand on time and attention in an environment where both are really limited.
As I’ve said, I don't see this as a new cycle. Agriculture has spent years being told to digitise, to adopt platforms, to integrate systems, that it's behind. Much of that effort and anxiety has still not really delivered what was promised. In many cases it has added friction, cost and more anxiety rather than removing it. After years of talking about data etc. Not much still gets collected. What does still isn't well utilised. Systems do not talk to each other. The burden of maintaining the technology sits with the farmer rather than the provider. There’s little in the way of structured help I could keep going for a long time...
The way I see it, Agriculture risks doing this loop again with AI. If the entry point is training, capability building, and data preparation, then the industry is once again asking farmers to do the work upfront. It assumes the problem is a lack of skill. In practice, the problem is that the tools are still young and frankly still a bit crap. The system is still not working as it should.
You can see the contrast with technologies that have succeeded. Auto-steer reduced fatigue and improved accuracy from day one. Variable rate made input decisions more precise and saved money. Tank sensors cut out time, cost and improved safety while giving peace of mind. The value was immediate and visible. No one needed to be told to adopt them, that they would take time to bed in and add value. They made sense in the context of the job. Interestingly to me they all also involved something in the real world - devices as well as just software (Maybe thats for another post.)
AI, especially generative AI does not yet have that obviously valuable use case, at least not as the main show. Instead, I think it’s being framed at a scale that still feels disconnected from day to day work. It’s being hyped as the next hotness. Talk of replacing jobs, restructuring the economy, building massive infrastructure. When that is paired with tools that feel inconsistent and unfinished, it does not build confidence and fuels anger and frustration.
The response from the research and innovation system to date has been to focus on skills and governance. Organisations like CSIRO are pushing capability uplift as the path forward and adding responsible AI as friction. Train people, build understanding, drive adoption, do it while jumping at your own shadow but that approach assumes that once people understand AI, they will use it. It ignores the possibility that people already understand the thesis well enough and the hype cycle well enough to remain unconvinced, and that they have enough interest to take specific time out ahead of everything else to upskill in the first place.
I hate mother statements like these but "farmers don’t avoid technology because they cannot learn it". They make decisions based on whether it improves their operation and provides value for their time and money. Just like anyone else would. Telling them they need to farm differently or take on board yet more information or develop more skills because AI exists is unlikely to change that. It adds to a long list of expectations that already put weight on the industry. It risks reinforcing the sense that decisions are being moved elsewhere and pushed down without regard for how work actually gets done or the context of people and place. Doing it without considering the system and as we all know there are some pretty serious system problems in agtech.
Personally, I believe we probably need to slow down a little, collaborate a little more and show a bit more humility to speed up. I don’t think the effective path is complicated, but I do think it is harder to execute, especially in the cacophony of hypers and consultants looking to make money (to be fair I am one of the consultants but I hope I think a bit differently). It’s hard to build systems that fit into existing workflows, that deliver outcomes that are obvious without explanation and remove effort rather than adding to it. Even more so when everyone is peddling the emperors new clothes. Agriculture needs this. Not more screens, not more interfaces that are ill considered and look like they’ve arrived from 1995 and certainly not as another report or dashboard.
Until that happens, I think continuing to push AI through training, consulting, and data readiness programmes is necessary but unlikely to produce the outcome the industry is aiming for. Especilly if in the next breath we're telling people junior jobs are going away (they're not). It will look like progress in RDC reports and strategies, it will look like policy from those trying to shape it and look knowledgable but on the ground until we get it right, it will feel like more work and expense for uncertain return.
That is not a skills gap. It is a failure to do the work to build the system needed to create the environment that will shift the needle. A field of rakes that agriculture, perhaps with learnings from agtech, big data and precision ag is placed ahead of other industries to navigate. Just without quite so many smacks in the face this time around.