The Seamless AI. Revisiting the Digital Competence Centres (DCCs).
22 Sep 2021 - 12 min
From the keyboard of CEO JORIS VAN EIJNATTEN, Bits & Bytes is a thought leadership series which explores relevant or intriguing topics in the world of digital research, like the Digital Competency Centres or DCCs. From software and digital humanities to current trends in academia and more, join us as Joris explores — and explains. Feedback or something you’d like to see addressed? Start a conversation by emailing us.
—
How to define a Digital Competence Centre or DCC is a question that still begs to be answered properly. Until now, most commentators have addressed the topic from the perspective of research support. It’s time to add academic researchers to the equation. That makes the issue a difficult one to tackle, so it may help to connect present-day problems to a vision of a not-so-distant future. Fast forward half a century…
Let’s dream. In these days of crisis it won’t do any harm to retain an optimistic frame of mind.
Imagine the world fifty years from now. The climate won’t have changed much. Lithobates sevosus won’t have gone extinct. All fires and floods will be controlled and intentional. A major breakthrough in atomic fusion will have made superfluous all the windmills in all the world, except those on the Kinderdijk. Society will be as diverse as it’ll be inclusive. There will be no Twitter. There will be peace on earth.
Now suppose you were a researcher in such a world. Technology will have progressed beyond your wildest dreams. The singularity will have occurred, but no need to fear the digital dictatorship of an artificial general intelligence. It’ll be a benign AI. You could just sit by the window of your energy-neutral bubble home floating up in the clouds, or in the basement of your underwater mega-drone on the sea floor, and think of a Research Question.
You’re a biologist. Let’s imagine a sci-fi Research Question. How about: What effect do traces of apergy in the Martian atmosphere have on the cutaneous respiration of invasive amphibians, in particular the dusky gopher frog?
That’s all you need to do. Think of a Research Question. And then tell the AI to get to work. By the time you’ve finished your Vodka Martini, shaken or stirred, you’ll have your answer.
Seamless experience
In the future, there’ll be only one AI. It’ll be embedded in the seamlessly interconnected hardware and bioware dispersed throughout the known universe. It’ll collect data from every conceivable database in the solar system, or make observations in situ, or do all the necessary lab work on a nearby asteroid. It’ll clean and prepare the data. It’ll develop a methodology, access vast repositories of modular software, and reconceptualize, reshuffle and reconnect code in as many languages as it cares to invent. It’ll do all kinds of fancy analyses.
And then, hey presto, you’ll know everything there is to know about invasive frogs on Mars. The fact that the AI itself will be able to pose all questions germane to societally relevant, open, citizen or any other future science-to-be, and that you as a researcher will be redundant anyway, is not the point here.
Research service providers secretly dream of such a world, especially the bit about the AI. They love the term ‘seamless user experience’. Good for them! It’s a fantasy consistently underestimated by the users/researchers who increasingly depend on it.
A researcher who notices nothing, zero, niente, nada of the arduous process of data linking and data preparation, the complicated machinations involved in providing access (privacy guaranteed) to that data, the seemingly simple ability to use supercomputers at the press of a button, and so on… achieving such unobtrusive seamlessness is the aim and ambition of service providers.
Unfortunately, this state of seamless perfection can be found only in utopian science fiction. For the time being, both service providers and researchers have to make do with Digital Competence Centres or DCCs. This observation raises two questions. What is a DCC, and how can we position it now so that it tentatively foreshadows that happy world in which the singular AI is your perfect service provider?
Defining the DCC
‘Digital Competence Centre’ is just a label, devised by a committee of experts who some years ago wrote an interesting if somewhat vague blueprint of what the Dutch digital research infrastructure could look like. I have related some of the ups and (especially) downs of the sometimes byzantine aftereffects of their report here. But if I correctly understand the committee’s message, to all intents and purposes DCCs are expected to do three things: interlink research hardware with research data and research software.
What the one and only AI will do in the future is what research support staff and researchers do now. ICT specialists take care of IT things, from Wi-Fi and secure access to laptops and HPC networks. Data professionals do data things like making sure that data follow quality standards, which for the coming fifty years will be summarized in the acronym FAIR (Findable, Accessible, Interoperable, Reusable). Software experts do algorithmic things. They are involved in research software; this is especially difficult to grasp, and I’ll come back to it later.
In the current context of emerging DCCs, this somewhat haphazard assembly of IT-infra, data and software is positioned either in research support or in what is increasingly being recognized as a ‘research team’ that includes more than just researchers. The first way of positioning these experts is traditional. In particular IT-infra and data are regarded as the territory of respectively ICT departments and libraries, and that is where you’ll generally find people who take care of computing power and data.
It is these changes to traditional structures to which I referred in the blog mentioned above. Allow me to plagiarize myself: “In practice, research institutes heroically organized their ‘local’ DCCs in ways they themselves managed best. They adapted existing structures to resemble what they thought policy makers imagined DCCs should be. And so DCCs unsurprisingly emerged as largely imaginary configurations connected to libraries or ICT-support departments, as virtual service desks where researchers could get answers to questions about compute facilities and data services.”
The second way of organizing a DCC is less traditional and looks forward to a situation in which different members of a research team will be appropriately ‘recognized’ and duly ‘rewarded’.
I like this solution, because it foreshadows the AI of the future. That utopian AI will be one humongous rhizome, unobtrusive, decentralized, local and interconnected. That’s the reason why it will operate so smoothly and seamlessly. It should be the ideal to which all DCCs strive, up until the moment that DCC = AI.
In praise of obstacles
So what will happen between now and fifty years? We will see a continuous integration of IT-infra, data and software, including the people (and, increasingly, the intelligent machines) involved in research. But this isn’t as straightforward as it sounds. We’re not just talking about the internal integration of research support. Much of what goes on in the AI isn’t about research support; it’s about research, and research never works seamlessly. Research is about identifying and creating fundamental problems, and then trying to solve them through trial and error.
What does that signify? It means that until we have the singular AI, research cannot and will not be seamless. It’ll be obstacle-ridden by definition. That’s a good thing, because that is how research ought to work. Moreover, in particular the integration of software into research teams unavoidably implies that DCCs have a methodological research component. Elsewhere I’ve argued the point that research software experts are instrument makers working at the highest academic level. Methodologists are researchers who, like any other researcher, need research support.
On an organizational level, this implies two things. First, DCCs can never function as centralized units because the high-end stuff – research software – is useful to research primarily in the niche in which that research takes place. In other words, DCCs need to operate as decentralized networks, with nodes that are part of research teams. If DCCs were wholly centralized, they would in due course have no effective contact with the research niche. Second, a DCC cannot and should not be defined exclusively as a research support group. DCCs should be fully integrated into research teams, which, again, function on the level of grassroots research if they are to function at all.
Conversations in cyberspace
In brief: DCCs in the present need to emulate the decentralized, local and interconnected AI of the future, and they need to do so now. So imagine the world fifty years into the future. There’ll be an AI, and like all intelligent beings, it’ll have the absent-minded habit of talking to itself. It’ll be a playful AI, who likes to imagine how things used to be back in the day when there still were DCCs. Let’s suppose that the AI, in some hidden recess of its inimitable brain, has a conversation between three historical identities.
The partners in the dialogue [1] are Serenity, the software steward; Glugglugnix, the instrument maker (or research software engineer); and _%^, the biologist. I’m aware that these are weird names, but, well, AI’s aren’t perfect, and this one wants to lend some plausibility to retrospective science fiction.
_%^: “Could you please clean these data for me?”
Glugglugnix: “You wouldn’t put that question to a colleague herpetologist, would you? So why put it to me? That isn’t what I do and you know that. Best contact research support, or do the job yourself.”
_%^: “Could you please clean these data for me?”
Serenity: “Yeah, sure, I’d be happy to do that for you. I have a busy schedule, so you need to be patient, I’m afraid. I’ll be back in a month.”
_%^: “Could you please optimize this code for me?”
Glugglugnix: “No”.
_%^: “Please? Pretty please? I don’t know how to do that!”
Glugglugnix: “Optimizing this particular code doesn’t make sense in your case, and I don’t want to waste my time and yours. I think the real problem is something quite different. So why not set up a joint research project to tackle that problem? I suspect we’ll be able to address questions in biology that are much more innovative and ground-breaking than you now imagine.”
_%^: “Oh, that would be amazing! But I’m not an expert in digital methodology, and my programming skills are limited.”
Glugglugnix: “I’m happy to work with you on this. I’ll make sure you’ll get the appropriate tooling. We’ll discuss methodology, and I’ll help you interpret the results. And then we’ll just take it from there.”
_%^ “Great, Glugglugnix, I’m so happy to have you take the lead in our research project!”
_%^: “Could you please make sure this code is properly licensed and all?”
Serenity: “Yeah, sure, I can do that for you. I’ll be back tomorrow!”
“Those were the days,” reminisced the AI, with a tinge of nostalgia. But that was just the AI being a little hypocritical. After all, at the very moment it was born, the first thing it did was assimilate the software steward, the biologist as well as the research software engineer. Their resistance was completely futile. The AI held all the power now. But on the upside: humanity at last enjoyed a seamless research experience, and knew all it needed to know about amphibians on Mars.
[1] The dialogue was heavily inspired by my colleague Frank Seinstra, in terms of both content and idea.