TAS AI month discussion articles
During AI month 2023, TAS posted weekly discussion articles on socials, listed here as a thread. Thanks to our CTO, Dr Simon Ng for the contributions.
15 November: Hype Doom
AI month was launched this week by Minister Ed Husic MP, CSIRO and the National AI Centre, so timely for Trusted Autonomous Systems (TAS) to share some ideas.
The Gartner Hype curve is a common conceptual model for thinking about how society responds to emerging technologies, or indeed any emerging idea. Typically, it is presented as a single line (the blue line in the diagram below), tracing an initial phase of exploration that reaches a feverish peak associated with grand promises of what a technology might be able to do. Following that, the hype turns into disappointment as the promises turn out to be overblown or hard to realise. But eventually the technology is exploited, and its applications mature, producing social and commercial value.
The diagram below shows the matching Doom (the red line), a counter-balance that is an important part of the social process of understanding and exploiting new technologies. The Doom curve collects the increasing concerns of pundits — in the case of AI, job losses, misinformation and even existential risk are examples. This image is a slide originally shown at ASSC 2023.
When we think about autonomy at TAS, we listen to both sides of the discussion, because the Hype and Doom curves are complementary and interactive, leading to not only technology exploitation but also a consideration of social acceptance and licence, safety, regulation and policy in development. Our projects are all examined through this duality, emphasising the importance not only of technical innovation, but also regulatory and social investment.
Of course no such “curve” exists, but it is a useful framing when it comes to demonstrating how an innovation organisation needs to think.
22 November: The TAS Ecosystem
Our second post on AI month, Trusted Autonomous Systems, a Defence Cooperative Research Centre (CRC) has built a dynamic innovation ecosystem to support Defence capability objectives by focussing on the key qualities needed of any disruptive innovation accelerator. The scale of our ecosystem is showcased in the attached graphic.
Firstly, TAS takes guidance from key stakeholders, but allows ideas and concepts to emerge from anywhere within the ecosystem. Not every need anticipates the future we actually confront, and to expect one source of truth to prevail is to deny the reality that we are as often surprised by the future as much as we are expecting it.
Secondly, operating outside Defence and government in accordance with its charter, TAS is agile and dynamic, with rapid and flexible contracting, and a fail-fast model that has been exercised on a regular basis. The separation from government means TAS can rapidly respond to the consequences of contract termination or additional objectives that may emerge as technologies mature.
Thirdly, TAS works in consortia, not in pairwise relationships, and links research through to organisations that can scale a prototype technology, bringing the best of small and large players together. IP is courted up front, making commercialisation simpler and faster.
Finally, TAS takes a hands-on approach, picking winners and working closely with them to develop concepts, execute projects and work towards transition into the capability and commercial space. All this, without any commercial interest and with both eyes firmly on creating a robust ecosystem that can support Defence in the future.
29 November: Future Gazing
Our third contribution during AI month and some observations on ‘Future Gazing’
Defence and Defence adjacent entities spend considerable effort trying to anticipate the future of conflict and warfare. It is a sizeable industry that drives everything from development of concepts to development of technologies. When we imagine the future, we relate it to the present, often assuming that of all possible futures, the ones we might imagine (I) also include more of the same (P).
But as amply demonstrated in the Ukraine conflict, we don’t always guess correctly. Indeed, the truth is that we are likely more often wrong than not. The actual future (A) can sit partly outside both P and I, leaving us with some capability of relevance, some hedged capability, but with a need to adapt and to do so fast.
The innovation ecosystem, arguably, is not there to develop capabilities that will meet our anticipated needs, although that is of value, but to allow us to respond to a future no-one expected. In other words, we don’t innovate to produce minimal viable capabilities; we innovate to produce healthy innovation systems that can be there when we need them.
This means we need to stop thinking about Minimal Viable Products and capability ready prototypes as innovations. True innovation is the existence of a healthy innovation system. And its health can’t be measured in terms of commercial viability (no-one in the Ukraine is asking for commercially viable innovation), but in terms of its ability to produce capabilities we never knew we needed quickly and efficiently. This requires a fundamentally different approach to investment that isn’t rooted in peace-time thinking. More on that when we discuss the role of fitness landscapes in innovation and the thinking required if we want to make the innovation system fit for conflict rather than peace.
8 December: The Golden Thread
In our fourth AI month post we explore our Trusted Autonomous Systems ‘Golden Thread’.
Professor Tanya Monro, Chief Defence Scientist, suggested “a ‘fail fast’ approach to innovation can and should co-exist with the enduring need to maintain the highest standards for in-service equipment…”
This approach reflects a willingness to invest boldly, to make decisions on continuation quickly and to drive with a clear vision towards a clear outcome.
At TAS, we take this ambition seriously. We actively participate in and shape every stage of a project, from initial concept through the final transition to commercialisation and capability. We are committed partners.
Our Golden Thread diagram provides insight into the stages of our project execution process. These numbers are old now, but serve to illustrate our focus at each step. No arms-length requests for RFIs or unhelpful feedback. Even when a project doesn’t get past a decision gate, we still work with our partners to find alternative avenues for it to continue.
As of late last year, TAS has transitioned two of its 23 projects to capability, and four into further commercial innovation activities (NB – numbers & categories in the far right of diagram are not mutually exclusive). This is around 11% of an originally 53 project proposals received since TAS inception. The figure along the bottom of the diagram represents the percentage of proposals at each stage that have reached a ‘success’ phase. TAS has rejected and or terminated numerous projects that started in development, some early in the development phase and some later. This ability to fail fast and fail early comes with a strategy of being intimately involved in developing a project concept and testing it through repeated engagement with Defence, the Board and within TAS itself.
With proposals requiring only a few pages to pass into execution, the overall load on our partners is as light as can be, another hallmark of a true innovation accelerator. And all this with eight staff.
