With all the recent hype around generative AI, ChatGPT etc, we at mOSS have been pondering about the real benefits that software driven systems and devices can deliver to humanity at large in areas that are really critical for its survival.
We are not talking here about the latest luxuries that the auto industry can build in to its executive model vehicle fleets, nor even are we thinking further ahead to the possible societal advantages of autonomous cars, buses and trucks. Though such benefits may in fact result, they are still too difficult to evaluate and it remains uncertain when and how the tipping point for some of the most heavily touted benefits (e.g. environmental amelioration due to reduced combustion engine CO2 emissions) will occur. Besides, many other technologies are vying for consideration in this field, meaning that the true societal benefits may only manifest in transportation safety and convenience if at all.
No. We mean something more tangible by way of benefit and/or something likely to bring greater well-being to those chronically under-served people in the world today: in short, to those communities whose needs are generally overlooked or side-lined due to the absence of a unified or unifying voice and consequently with little leverage in our modern capitalist economy.
A brief sidebar here: this is not a theme that is alien to the technocrats of today regardless of their political persuasion. We hear it over and over from the proponents of 5G both inside and outside of government. They refer to it as narrowing the technology divide or somewhat more positively as the democratization of technology.
The first area we looked at was healthcare.
The thing to mention right off the bat is the safety factor. Based on our significant experience with the automotive industry, we have learned that it is too dangerous to introduce open source software into safety-critical systems such as engine control units. While vehicle manufacturers have been happy to open the doors to OSS in other parts of the vehicle infrastructure; consoles, chassis systems etc., ECUs have usually been designated as a no go area.
As in automotive, so also in healthcare one would suppose. Critical safety malfunction could forseeably result in death or serious injury. The presence of OSS might introduce software vulnerabilities potentially increasing the risk of a denial of service cyber attack or worse. So the argument goes. Incidentally, we tend to ignore the fact that these risks are equally present in all software whether open source or proprietary.
So, imagine our surprise to learn that OSS is being used, and being used successfully, in insulin delivery systems for the management of diabetes. According to an associate professor of endocrinology at one of this country’s most prestigious university hospitals, the customization advantages of OSS and the inflexibility of the market leading industry providers of proprietary systems, has sparked a movement within a movement. Evidently many diabetes patients have grown tired of waiting for their individual needs to be met by big tech. Due to its infinite capacity for nimble customization, OSS would seem to lend itself perfectly to such a use case.
Diabetes is prevalent enough today that the subset of the population with the disease is a faithful reflection of society at large, meaning that the technical abilities and software development skills are present in large numbers among this community. There is absolutely no reason to think that the software product being generated by one of these diabetes open source communities should be in any way inferior to their counterparts in the world of proprietary tech. The ingredients of drive, autonomy, commitment and user engagement are typically what propel open source communities to success and longevity.
Still uncertain whether this story was a one-off rather than a real life proposition and with lingering doubts about safety, we dug deeper.
In November 2021 a paper was published by the leading medical journal The Lancet entitled “Open-source automated insulin delivery: international consensus statement and practical guidance for health-care professionals”1.
The paper is illuminating for many different reasons, not least in respect to the challenges that open source software presents to a medical profession largely ignorant of the concept and possibly influenced by some of the highly publicized and derogatory remarks made by certain major players in the software industry. We should not forget that both the open source movement and the open source code that it generated was frequently referred to as a ‘virus’ capable of ‘infecting’ not only our computer systems but also our business values and models. These lurid labels came about not as a result of any inherent defect in the code itself but rather because of the effects of a particular group of licenses used by some of the open source pioneers and the attempts of the legal profession to describe a phenomenon they had not encountered before. By and large this negative profiling has been assigned to history… but mud tends to stick.
The findings are generally positive in favor of this globally developing patient-led OSS initiative, which being open and transparent is providing an ever increasing pool of real-world data and user-driven evidence around safety and efficacy. Safety concerns are certainly not brushed aside by the authors but they are addressed in a measured and proportionate way in the context of the overriding ethical duties of medical professionals and their representative bodies to fully and properly inform their patients (and of course themselves first) of all available and effective treatments, including “patient-driven, open-source innovations”. With regard to the latter, emphasis is placed on the need for support and guidance in clinical environments.
Important recommendations are given for all key stakeholders involved in diabetes and associated technologies, including the medico-legal profession and regulators. In the face of an apparent reluctance among powerful regulators such as the FDA to give equal status to proprietary and open source solutions, the paper concludes that “problematic regulatory positions” should be challenged and clarified.
It seems that in this case at least, industry power and state regulation is not going to stand in the way of the people.
The second area we looked at was agriculture
There are a growing number of national and international initiatives in this area, with the declared objective of creating open platforms for sharing useful and innovative technologies among a global network of farming communities, producers, agri-science and agri-data researchers, cooperatives, food distributors, government regulatory bodies and standards and certification agencies.
Given the vast regional and local disparities it is both logical and critical that access to core technology (e.g. the various levels of farm management operational software and applications) is made universal and with minimal barriers to entry whether in the form of cost (e.g. license fees), practicalities (e.g. language barriers) or other trade-offs such as loss of protection of proprietary information and control over private data.
Equally critical it would seem is the need to take advantage of the opportunities provided by the open-source software model to build communities and eco-systems, in which customization of specific solutions can flourish and digital products can be tailored for the wide diversities of climates, regional and local ecologies, soil qualities and soil management techniques and any number of additional factors involved.
The scale of these endeavors is potentially huge and as in the case of the diabetes community the opportunity for collecting and consolidating real-world data could have far reaching implications.
Clearly, for these initiatives to succeed in terms of establishing a high degree of global digital parity and at the same time enabling regional and local producers to adapt new technologies and capitalize on innovation, it will require considerable aptitude, commitment and cooperation among those at the helm as well as those at the grass roots (no pun intended!).
Statistics indicate that the use of open source software in the agriculture industry as compared with an array of other prominent industry sectors is relatively low. Whether this translates into a generally lower level of OSS awareness and OSS management maturity is hard to assess but some analysts have made this connection. With the advent of AI techniques and tools, the already existing complexities and challenges have taken on a whole new dimension. The temptation to rush ahead with AI and generative LLM products in particular must be close to overwhelming.
Delivering real opportunity for people in some of the most disadvantaged areas of the world is becoming something tangible and tantalizingly close either as a means or a by-product of these agricultural initiatives. LLMs are being used and adapted for example in Africa and India to give indigenous people and the languages they speak a new outlet and an exposure they probably never dreamed of.
In the meantime, however, we need to remember that much of the AI being produced today is still closed and as such is hard to evaluate from the perspective of both quality and legal risk. Some of the critical issues mentioned above may be subject to compromise unless appropriate vigilance is maintained.
There is so much at the end of the rainbow as long as we learn to walk the OSS backroads before we fly high into the blue AI sky.
Your mOSS team
1 The Lancet: Diabetes and Endocrinology. Nov 13 2021 Volume 10 Issue 1
Leave a Reply