As it weaves its way through our lives, AI has become the subject of intense debate in university circles in Australia. However, the focus of this debate is yet to fully align with the new reality we're stepping into. A paradigm shift is happening beneath our feet, but the conversation hasn't caught up.
Much of the discussion around AI in the education sector thus far has been centred on its implications for teaching—how to circumvent AI-assisted cheating, how to enhance plagiarism detection, and how to manage the occasional yet disconcerting generation of false information or factual errors. These discussions overlook a critical facet of AI's potential in the university setting. A far-reaching, transformative opportunity lies in harnessing AI for research—a realm where AI has quietly been making headway for some time.
AI introduces not just an assistant or a tool, but an active participant in the scientific discovery process. We're witnessing a revolutionary shift to AI as a co-producer of knowledge, a change that has seismic implications for the scientific world. This paradigm shift—under-acknowledged, yet resoundingly potent—stands to reshape the world of research as we know it.
Are we ready to embrace this new paradigm and take full advantage? The answer could very well determine the future trajectory of Australian universities on the world stage if we can get our response right.
Changing Paradigms: From Information Retrieval to Information Co-production
The likes of meta-databases like Scopus and Web of Science, and the long list of academic databases, have been commonplace in academic research for decades. But this old order is changing, giving way to a new paradigm. The traditional model of database information retrieval (IR) and simple citation connections—a passive process of extracting pre-existing knowledge—faces obsolescence. In its place, a model of information co-production is rising, enabled and driven by AI.
The IR paradigm, although instrumental in the early stages of digital knowledge management, is now outdated. The model relied on human researchers to search, filter, and extract information from vast databases—a laborious, time-consuming process. The IR model was largely linear, failing to capture the intricate relationships and connections that bind the universe of scientific knowledge together. AI, with its capabilities of deep learning and pattern recognition, introduces a revolutionary shift from this status quo. Rather than focus on IR–is still the focus when we talk about factual errors in AI–we should focus on AI's ability to interpret, connect, and generates insight, thus co-producing knowledge and building broader 'knowledge cartographies'.
This paradigm shift is greatly facilitated by the advent of Open Access (OA) science. OA publications have democratised scientific knowledge, making it widely available and breaking down barriers to information access. This trend has significantly increased the quantity and diversity of data available for AI algorithms to process, learn from, and contribute to.
The Open Citations movement and large open databases like Lens.org further propel this paradigm shift. These platforms are committed to making citation data—traditionally locked behind paywalls—freely available to all. The OpenCitations Index of Crossref open DOI-to-DOI citations (COCI), for instance, hosts over 1.4 billion citations, representing a vast treasure trove of interconnected scientific knowledge. COCI treats citations as first-class data entities, providing a rich, complex resource for AI systems to extract, analyse, and build upon.
These trends are fueling a transition in scientific workflow software from IR, to information co-production. The universe of scientific knowledge stands on the cusp of an AI revolution, with opportunities for discovery expanding at an unprecedented pace.
Case Studies of AI as a Co-producer of Knowledge
The transformation from information retrieval to information co-production is not an abstract concept—it is already in play. We can witness it unfolding in groundbreaking applications across various domains of research. Let's explore three pioneering instances where AI serves as a co-producer of knowledge: Scite, AlphaFold, and DABUS.
Scite is an AI-driven platform that transcends the conventional citation index to offer what it calls "Smart Citations." These citations don't merely point to a referenced source; they provide context and categorise the citation as supporting, contrasting, or merely mentioning a claim. The AI examines the full text of scientific articles, extracting and analysing citation statements, and classifying them using deep learning. This provides a much richer understanding of how scholarly works influence each other, transforming the AI from a tool of IR to an interpretive tool capable of co-producing ideas based on interconnected knowledge.
AlphaFold is a revolutionary AI system by DeepMind that predicts protein structures with remarkable accuracy. Traditionally, determining protein structures has been a painstaking, expensive process. AlphaFold, however, employs deep learning techniques to predict these structures, thereby significantly advancing the field of discovery research. By contributing directly to scientific understanding in this way, AlphaFold represents a paradigm shift, as AI moves beyond IR to participate in the very heart of scientific discovery.
DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) is an AI system claimed to have inventive capacity. Unlike conventional machine learning systems trained to solve particular problems, DABUS devises and develops new ideas—akin to the human inventive act. Notably, it was recognised as the 'inventor' of two patents, one for interlocking food containers designed for robotic handling and the other a warning light that employs neural activity-like rhythms. By initiating new ideas and contributing to the invention process, DABUS illustrates the transformative potential of AI as a co-producer of knowledge in the field of invention.
The Future of Research Workflows: Incorporating AI and High-Performance Computing
Research is at a pivotal juncture. AI has transitioned from an optional aid to an indispensable partner in the pursuit of knowledge, demonstrating its immense potential in transforming research workflows. However, the true potential of AI in research lies in its successful integration and scaling across all scientific and research workflows, which is an area where there has simply been no investment to date naitonally or by institutions.
The integration of AI into research workflows opens up new possibilities for increased speed of discovery, broader collaborations, multidisciplinary integration, and the acceleration of knowledge production. But to optimise this shift, we need the horsepower to drive it. That's where high-performance computing (HPC) enters the picture.
Enter cloud-based HPC solutions, like Amazon Web Services (AWS). AWS and other platforms provide virtually unlimited computational power and storage capabilities, allowing researchers to deploy AIs tackle complex problems at unprecedented scales and speeds. These platforms facilitate the running of large-scale simulations, complex data analyses, and intricate computational tasks, essentially supercharging the research process.
By leveraging AI tools plus this compute power we can drastically hasten the pace of scientific discovery and innovation. AI and cloud-based HPC together hold the promise of an era where research is not just about finding answers, but co-producing them with AI at higher speeds, greater accuracy, and more agility than ever before. Embracing this approach could profoundly transform the landscape of scientific exploration.
The Role of Australian Universities in Embracing AI for Research
As we stand on the cusp of this revolution, Australian universities have a unique opportunity, and indeed a responsibility, to lead the way in integrating AI into research processes. Harnessing the power of AI for research is no longer just an innovative idea, it's a necessity for maintaining a competitive edge in global research.
Unlike in the past, in this future the challenges of scale posed by our population constraints need not be a barrier to the impact of our research. The adoption of AI and associated technologies in our research workflows can provide a counterbalance, allowing Australian science to finally (legitimately) punch well above its weight on the world stage. Through the clever integration of AI and high-performance computing technologies, we can achieve a level of research output and influence disproportionate to our population size.
Adopting AI across our research workflows would allow for accelerated discovery, more efficient coordiation and of use of our limited resources, and a more profound understanding of complex phenomena, positioning Australian science at the forefront of global innovation.
In this brave new world of AI-assisted research, Australian universities should not only participate but strive to become leaders. Embracing this paradigm shift will be critical in ensuring the enduring relevance and impact of Australian research, shaping a future where our universities are recognised as global powerhouses of scientific discovery.
The opportunity persented by embracing AI as a co-producer of knowledge marks a seismic shift in the way we approach research. As we transition from traditional IR systems to dynamic AI-enabled information co-production, we are not only changing the tools we use but fundamentally transforming the nature of research itself.
This paradigm shift, facilitated by open access science and the wealth of data available through open databases, allows us to harness the power of AI in generating novel insights and accelerating discovery. By integrating AI technologies across our research workflows, we have the opportunity to revolutionise the scientific landscape and propel research efforts to new heights.
In the context of Australian universities, the implications of this transformation are particularly profound. Far from being an insurmountable challenge, our relatively small scale could actually be a strategic advantage in rapidly adopting and integrating these powerful new tools in a coordinated national approach. With a bold vision and a proactive leadership, Australian universities have the potential to become global leaders in AI-enhanced research.
This pivotal moment in scientific history calls for bold action, and there's no better place for this evolution to occur than Australia. By embracing the paradigm shift toward AI co-production in research, we stand to gain a world-leading edge. It is incubment on our leaders to make this happen.