I could use up acres of the blogosphere in detailing my deficiencies as a blogger. But one deficiency is particularly applicable here. I do not by and large, place too much faith in polls or surveys, especially where the identity of the respondent is known to the organisers. I have entertained these suspicions for a long time: they were the subject of my first published letter in the Guardian newspaper in 1964. I had been working as a Gallup pollsfer and had begun to appreciate how responses to my questions varied with the reading made by the respondent of who I was and what my expectations might be. If this could happen as I patrolled the pavements ofthe Angel, Islington, I reasoned then, how much more likely is it to happen now . To happen, for example when, research academics respond to the latest survey on the Towards Responsible Publishing proposal promulgated by cOAlition S.(https://www.coalition-s.org/coalition-s-welcomes-the-findings-from-the-consultation-on-the-towards-responsible-publishing-proposal/)

Let me hasten to add that this is not because I think that researchers would falsify their position in order to curry favour with the research organisations conducting the survey. I think it is more the case that the 11,000  people responding to the survey would have a clear idea of the probable direction of cOAlition S policy and would be anxious to help drive it forward while those in disagreement would be less likely to respond at all. I do not say this to denigrate the cOAlition S direction, since I have a great deal of sympathy with their conclusions. And I largely agree with most of them. However, my experience of efforts to talk about the future has led me to different conclusions. my belief now is that we get a better fix in the future if we look at The working problems of current information distribution in the context of networked ecosystems over the past 30 years. The issue is not so much the practices of publishers; network communication will redesign itself around the critical dissemination needs. in the history of these changes, OA (open access)will be a really interesting historical footnote. But aa footnote all the same.

And then I encounter another difficulty. Revolution : what do we mean when we describe change as revolutionary and would we recognise a revolution if we saw one?. In watching the development of publishing in the scholarly research field, as a participant and as a advisor and as an observer for well over 50 years, I conclude that usually when we hail a “revolution“ it is a hoped for event rather than an actual one. And that the real revolutions, meaning far reaching or complete changes in systems and services, happen quietly and usually take a little retrospective time for recognition. So it is with  OA (open access). The revolution here was “free to me as a researcher at the point of use“. I am confident that this will never be reversed. whatever the arguments about the level or affordability of author publication charges.The issue of how this is paid for, such a momentous issue for publishers, is not a great issue in the larger picture of research funding as a whole. As a useful corrective think through the issues from the point of view of a senior official responsible for scientific research in the Indian or Chinese governments. China today and India in a decade will be the major sources of research articles: will they be creating major budget for publishing charges?

I think also that, starting in Western Europe,we are seeing the slow disentanglement of research article productivity from issues of pay and preferment. here it is the USA which seems to be the exception. i have also experienced the anxiety of researchers and academics, particularly in medical research in recent years, to use the corpus of published research from their institution to attract more researchers and more funding.

And then people , especially in physics and life sciences, tell us about the other great problem. Speed. Getting it out there while the issues are hot, and while the research team and its institution gain kudos from being in the argument and feeding it with fresh evidence and new discoveries. Apparently something like 10% of research articles published last year first appeared as pre-prints during the year. I think this can only grow, and when we reach, in the sciences in particular, the position where most articles come from pre-print servers initially, we shall see, as well, an impact on journal branding and it’s importance.

I do not see the branding of the great journal names altering at all: Nature ,Cell, Science are in a very safe place. But when, as I read recently, a communications letters journal publishes 21,000 articles in a year, it is not so much a brand as a channel. I see a potential consolidation of brand to institutions. In an age where all publishing of this type is digital, all processing can be effectively accomplished in AI moderated environments: when peer review can be semi- automated as well , subject to human supervision and approval, then it is surely not hard  to imagine that the research institutions will want to secure the branding and represent their corpus of knowledge and their productivity in one place? . And publishers of today will surely become their collaborators with this tomorrow. those who are despondent about the disintermediation of publishers and publishing should cheer themselves up with the huge prospects offered by collaboration with research institutions on one side to create secure branded repositories for articles and attached evidential data, while on the other side looking at the intelligent commercial reuse of such data in industrial and commercial AI applications. this is not the time for publishers to have a Gutenburg moment and retreat to the scriborium

Anyone still looking for revolutions? Try the revolution in article reading – some now  believe that more articles are initially read by machines in the sciences than by people, and others assert that this happened at least 10 years ago. At the same time it is clear that the volume of research could not be handled at all if machine based summarisation was not available. This makes the idea that some publishing majors are still not employing metadata which facilitates machine to machine communicationtotally bewildering. Perhaps it really is time to accept that all research has to be free at the point of academic use, forget OA, and settle down to working on the real problems of the research ecosystem in the 21st century..?

 

What happens when you are in the middle of a revolution, but nothing radical appears to be happening? Richard Wal

ters, in an excellent article in the Financial Times (11 July 2024) speculates about the US tech bubble, and the potential for rapidly growing valuations to come into conflict with apparently little deployment of AI based products and services at scale. My friends and colleagues at Outsell reflect a trend by talking about AI and the hype cycle, with generative AI now moving from the peak of, inflated expectations into the trough of disillusion, presumably prior to ascending to the plateau of productivity. But we are still talking about AI everywhere and in everything, and personally, having experienced AI in previous guises as neural networks or as expert systems or as robotic process automation, I have to wonder if we are thinking about change in the right way.

What if change is not revolutionary, but osmotic? What if we should be thinking not about startling use cases but about the rate of general diffusion of AI everywhere? Is assimilation less transformative than step change? When I wrote recently about the applications of AI in Apple Intelligence, I realised of course that this is an announcement of a process that  has been going on for many years and which will go on for many years: adding more utility progressively in each new release and iteration of existing software-based products and services. It’s not very exciting, is it? Even an Apple Press release could not make it sound revolutionary or radical. It certainly does not tune in nicely with the way in which stock markets work: This is not the sort of gold rush investment cycle that we seem to require in speculative terms. Yet it does seem to be what has happened in all of those other AI investment cycles since the 1980s. There have been disappointments on the way, of course. Think of Autonomy, once boosted as breakthrough technology, ending in endless disputes in the courts over whether its radical qualities had been misdescribed. There may well be disappointments of a similar sort in the future: some of the development programs of generative AI market leaders are “too expensive not to deliver “. Or at least, the companies concerned can never admit comparative failure. Again, there may be scale issues. Perhaps small learning models (SLM) will be more useful than the LLM research that enabled them. Perhaps AI in niches will be revolutionary (collating and reading test results in radiology and MRI scans) but related technologiesmake very, very slow progress towards universality (driverless cars). Whatever is the conclusion of history when all this is being reviewed, one thing, I suspect, will be clear. We would never have got anywhere without the huge concentration of investment in research, even if in retrospect we can see how much we wasted and how differently effects were from the intended benefits.

I do believe that we will look back in 2030 and see that fundamental, radical and far reaching change did take place in the mid 2020s. I think that if we compare 2020 with 2030 we will see that machine intelligence has transformed the way in which we live and work. But if we look at 2025 from 2024 and try to anticipate that change then I think we will see something different. We will see continuing slow progress with consumer applications led by companies like Apple, gradually enabling voice driven systems to free users from screens and keyboards and enable the more fluent and transparent use of technology. I also see that corporate of all types will gradually automate systems and services to create greater productivity, shed manpower, and understand users and markets with greater precision. The new products and services however may take longer. In the information, data and analysis marketplaces in which I have been most concerned for the past 55 years, we will see the gradual infusion of machine intelligence so surreptitiously that many of us will not notice what is happening. Make no mistake: from this point on we will all describe everything we do as “AI powered “or “AI based “, but that is simply the world of slogans. In truth the world will change, to coin a phrase, “not with a bang but a whisper! “

« go backkeep looking »