Mid-March of this year saw the very first EverythingOpen conference, held in Naarm/Melbourne at the Melbourne Exhibition and Convention Centre. Unfortunately, I had some PhD commitments, and only managed to get there for the conference dinner, but was able to catch up afterwards via the conference videos – and this blog post is a recap of my favourites.
Why EverythingOpen?
EverythingOpen is a re-imagining of linux.conf.au. linux.conf.au – affectionately known as LCA – started life in 1999 – as the Conference of Australian Linux Users (CALU), and has been a deeply-technical fixture on the Australian open source calendar since. There have been deliberate moves (which, for transparency, I strongly support) to widen both the audience of the conference, and the pool of volunteers which delivers it. EverythingOpen, the brainchild of long-term Linux Australia leaders Sae Ra Germaine and Joel Addison – aims to connect the many open* communities in Australia together.
There are open source communities – those who create software free for everyone to use, and enable that software to have impact (with thanks to Donna Benjamin for her work on enablement) by building communities, resources and other artefacts to bring software to life.
There are open hardware communities – Jon Oxer, Andy ‘Geekscape’ Gelme and Hack Melbourne – and their incredible work delivering the Open Hardware Miniconf at LCA for nearly a decade with projects such as the Swag Badge and the Dingo Car.
There are open data and open gov communities – Pia Andrews’ work here comes to mind in her founding of the GovHack movement, which now has a life of its own; Fiona Tweedie’s work with Open Knowledge, the work of Liz Stokes and Sara King and others who work with open research data and the researchers who need the skills to work with that data.
There are open security and open telemetry communities, bringing the ethos of open data and open source to the prescient problem of infosec and cyber security; Cameron Tudball’s talk in this space is definitely in my favourites list.
There are the open GLAM communities – Galleries, Libraries, Archives and Museums who often use, create or adapt open source tools into their curation and collection processes. The work here of Hugh Rundle, Ebe Kartus and the many good folx of VALA comes to mind.
There are open science communities – researchers, infrastructure specialists, high performance computing (HPC) experts – who increasingly use the power of machine learning and artificial intelligence to make new discoveries and advances – but who need a wealth of expertise in software, hardware and infrastructure to do so – the work of Kiowa Scott-Hurley
There are open education communities – educational developers, digital literacy specialists, knowledge management people – who create resources to educate, to inform, and to share learnings.
EverythingOpen is glue: helping these communities connect, collaborate, and converge.
Theme: Supporting our students, researchers and educational developers to create new knowledge and make discoveries (#OpenScience, #OpenData, #OER, #OpenEducation)
Liz Stokes, Australian Research Data Commons – Hammer Time: Building Carpentries community in Australia
Liz’ talk introduced The Carpentries – a volunteer, membership-funded collective which aims to teach foundational coding and data science skills to research communities worldwide, using a train-the-trainer approach, and engaging with the social factors of learning. In Australia, The Carpentries is supported by the Australian Research Data Commons, and the National Research Infrastructure Strategy.
As Liz explained, The Carpentries is a response to several trends:
- the increasingly computational nature of advanced research – with everything from linguistics, to pharmaceuticals, to physics, to art and design – now harnessing high performance computing (HPC) in the form of GPU clusters – to produce research outputs;
- the lack of basic computational courses offered in non-STEM courses – for example in the humanities and arts, graduates may never use tools such as Git, or Python;
- and the increasing challenging finding assistance with these tools within the academy, as budgets are squeezed – and communities of practices often don’t exist internally.
My key takeaway in this talk is just how much research infrastructure skills training takes place on a volunteer basis; universities are crying out for Carpentries volunteers – but with grad students taking on tutoring work just to stay afloat – who has time to take this additional workload on? An alternative model here would be for universities to fund Carpentries volunteers – but again, universities themselves are strapped for cash. It wouldn’t surprise me if startups spring up in this space – offering a similar “Git and Python bootcamp for $$$” style offering – which privileges these skills even further.
Another interesting aspect that Liz mentioned was the growing support for a new employment category within universities in Australia to support research software development. The skills required here often don’t fit neatly into the traditional “Academic” versus “Professional” dichotomy – instead straddling competencies of both – resulting in the inability to progress professionally in either.
Kat Cain, Deakin University Library – Open by design: co-designing and producing an OER interactive textbook
In this informative talk, Kat Cain, Library Partner at Deakin University, reported on the design and publication of an OER (open educational resource) textbook, which was designed using the Pressbooks platform. I hadn’t come across this platform previously, and discovered how it could be used to create engaging and interactive resources for courses that were being delivered in hybrid or online mode – which, given the pandemic, is likely to be a feature of the educational landscape for some time to come.
The aspects of Kat’s talk that I really enjoyed were the interdisciplinary nature of the project, and the challenges this presented – across methodologies, toolsets and in terms of project processes. The team adopted a Peopling Technologies philosophy, which was again something I hadn’t encountered before. In Kat’s words, this perspective holds that:
“People need to be positioned at the heart of any engagement with technology.”
Kat Cain, EverythingOpen
Dr Linda McIver, Australian Data Science Education Institute – Raising Heretics on a diet of Open Data
Linda has a background in computer science education, and her passion is in teaching kids about “data in the real world” – showing them how they can use critical thinking with real data to pique – and sustain- an interest in data science.
Linda’s talk highlighted some of the challenges of data science education – such as that late high school – when students are often first introduced to data science concepts – is too late to stop them dropping out of the STEM pipeline – and, moreover – that teachers have often never been taught to teach STEM – and are terrified at it.
Linda’s engaging talk relayed several examples of how students, using real-world data, were able to ask questions that had never been asked before – another proof point for the value of open data.
Theme: Representing our world in the age of generative AI (#OpenData, #OpenSource)
Peter Neish and Alex Lum – Sustainable open data using WikiData
This very informative talk by Peter and Alex covered the WikiData project – which started life as a way to create re-usable elements for Wikipedia pages across languages. WikiData creates linked open data through defining ontologies. What this means is that WikiData creates and connects structures to help represent our world. It can be used to map people – and their characteristics; places – and the places that are related to each other – and products – and the people who make or use them, and the places they’re found. These relationships have meaning – they are semantic – and can be queried to help provide answers to questions. Wikidata is an implementation of what’s known as the semantic web – a movement that’s over a decade old, but only now starting to gain momentum and recognition as its applicability to technologies like the metaverse, and augmented and virtual reality – becomes clearer.
The key implication I see from the WikiData work is actually in generative AI.
With the advent of generative AI, one of the challenges we have is to represent the relationships between objects in the world. While large language models like ChatGPT can scrape swathes of text from the web, and infer relationships in that text using attention-based transformer models, having a set of object ontologies – structures, hierarchies and mapped relationships – is a way of representing or validating relationships between entities in the world. For example, an LLM might infer that “Paris” is closely related, statistically, to “France”, but it doesn’t know that Paris is a city, that France is a nation-state, and that the relationship between the two entities is “is the capital of” (which also implies a one to one cardinality).
The challenge with LLMs at the moment is that they don’t have valid representations of relationships in the world. They can infer, statistically, what word comes next in a sequence, or generate plausible text, but we don’t trust the text they generate. Without real-world representations to validate against, we’re simply going to stop believing what ChatGPT and its kin generate.
And this is where the WikiData work has a huge role to play. By mapping relationships between entities, these can be used to validate the output from LLMs. “Is Paris the capital of France?” – using WikiData, and linked open data, the answer to this question can be validated.
Theme: Supporting our public service to deliver digital government services (#OpenGovernment, #OpenData)
With efforts in the Australian Public Service to create whole-of-government digital services in tatters with the already-poor reputation of the Digital Transformation Authority eroded further by a damning Auditors’ Report, there is a renewed opportunity to connect government efforts to broader open communities.
Pia Andrews – Creating open source legislation as code
Pia is well known in open government circles, and as a founder of the GovHack movement, and here she advocated for creating open source legislation as code – helping to build trustworthiness into government services. By having legislation as code, the legality and validity of government decisions can be verified, which in turn helps to ensure repeated, correct decision making.
You can read more about the concept of trustworthiness in government decision making in this paper, which I helped to peer-review:
@misc{andrews2022trust,
title={A Trust Framework for Government Use of Artificial Intelligence and Automated Decision Making},
author={Pia Andrews and Tim de Sousa and Bruce Haefele and Matt Beard and Marcus Wigan and Abhinav Palia and Kathy Reid and Saket Narayan and Morgan Dumitru and Alex Morrison and Geoff Mason and Aurelie Jacquet},
year={2022},
eprint={2208.10087},
archivePrefix={arXiv},
primaryClass={cs.CY}
}
Parting thoughts
EverythingOpen undoubtedly delivered on its promise – it started to connect our disparate, diverse open communities of practice together, forging new relationships, and hopefully, fresh collaborations. My key concern is whether this initial impetus can be maintained; Joel and Sae Ra are both long overdue for a break! Will another group of community practitioners pick up the mantle – and help ensure Everything Open remains a fixture on the calendar? I hope so!