Oct
14
I am on my way back from Boston. These days America seems a strange place, pivoting towards extremes and polarised as never before in my experience. It has become a land in which I am afraid to discuss politics. My American friends are still my American friends: charming, cultured, intelligent. I am still in love with them, but they do not talk politics either. We have grown afraid of what we might hear ourselves saying, and what effect it may have on the people to whom we are saying it. It was not like this when I first visited this city in 1974.
So it is best to concentrate on business. Over 100 companies, members of the Outsell community, met together in Boston this week. From my own viewpoint, as a rather ancient observer who has attended each of the previous 17 such meetings, I thought that this was a great success. It had that element of turning point, of mutual and collective recognition of change.Of course we had to get through the ritual of complaints about the state of the USeconomy first. A short spell in Europe would convince most people that the American economy is in a pretty healthy state. I do not want to downplay the problems, but in terms of growth rate and in terms of employment, the American economy is in an enviable state and it is clear that the US based companies that I have spoken to this week have been investing steadily in the future for the past year, a year in which margins have not touched record levels, but are still so significantly restored from the pandemic period that reinvestment is a realistic possibility. But any optimism here is held down under a heavy cloak of caution. It is easier to persuade oneself to invest in infrastructure or in cost saving retooling, or in headcount reduction through technology adoption than it is to invest in new product development for markets which are emerging but which do not quite exist yet.
At the previous meeting of this community , in 2023, attendees were given no doubt that the vision of the future was an AI infused vision: machine intelligence was going to drive the world of content generation from publishing and information provision through data and analytics and into a Brave New World of generative AI. This was an important moment. It was like saluting the Internet before the creation of the Web. Or saluting Gutenberg. without knowing that the scriborium had another hundred years of successful operations still to run. in other words, it was a moment of faith and belief and expectation. It was not a moment for concrete demonstrations and outcomes.
Yet recognising change and saluting it has never been more important. The 100 years that they waited from the mid 15th century for the printing revolution to actualise itself had their equivalent in the 14 years that we waited between 1979 (the launch of my first searchable database) to our first real Internet connection in 1993. Sustaining investment during the time period required, in this case,to move digital from dial up to TCP/IP was a real problem. As well as good sense and good judgement it also required, as investment so often does, faith.. And, as I have often said here, the road to AI has not been a short one. In my own terms I dated my own faith from a Media Lab seminar at MIT run by Marvin Minsky in 1985.
So what did I really expect of the 12 months between Outsell 2023 and 2024? I travelled to Boston hoping for more of the same.
More cautious optimism? More gradual exploration of the commercial issues ?What I did not expect was a room full of companies all experimenting in some way or another with the concepts, ideas and investments involved in utilising AI in some way in their business. I sensed a real change in business attitudes. I do not now see information based businesses sitting back as at the beginnings of the digital age, and waiting for something to happen. I do not see small businesses backing away from change, or big data-based businesses saying that they will wait upon big AI tech to sort out the world before they invest for themselves. The picture emerging now is complicated, but it is about engagement and about embracing change. Above all, in the information and communication sector, I sense an eagerness now to look hard at the technology and ask what it may do for them. This time round, rather than become the victims of change, the industry that I have served for the past 55 years seems determined to be an instigator of change. And that is a very welcome relief!
if you strip away all of the analysis and all of the strategy, investment at the final point of decision becomes a balance of hope and fear. Many of the people I have spoken to this week entertain the real fear that unless they can bring costs down and re-engineer their processes, whether they are creating a credit rating or a scientific research article or drafting a legal document, they are going to come up against competitive headwinds, as ungentle as hurricane Milton. As I heard of these developments from the platform and discussed them around the room, it was apparent as well that re-engineering process through AI was also about adding value , and thus increasing competitiveness.
The meeting this week contained a session on improving data quality for AI. Five years ago corporate response to this sort of presentation was negative: it was all very well to talk like this, but these things cost money and nobody was prepared to pay more for better data. Here the atmosphere was totally different. There were few players in the room who were not actively improving their own data, and no one was not aware of the role of machine intelligence in automating and accomplishing this.
.
Flying into Boston, my mind was full of Big Tech licensing controversies, the regulatory framework that is emerging and whether we were all going to be swallowed up by Big AI Tech, who would buy our data for a pittance and then “hallucinate “it into the search boxes of our users. These preconceptions were entirely wrong. Much more time was spent discussing case studies and exemplars. There was more emphasis on using small learning machine environments, RAG, and applications dependent upon company-owned data sets, plus some licensed files, and publicly available data inputs . Yes, you could create your own training sets for your own application in your own niche.. You did not need to ride on the back of the big AI developers:some were using open source or licensed AI software. The urge to experiment was borne both from a perceived need to anticipate competitive threat – “ they seem to think that anyone can come in here, buy the right data sets, engineer them to user requirements, and take our markets away from us “– and an optimistic feeling that the existing players were in a more powerful position than they had ever realised. They had the customer relationship , they knew the needs and requirements and the ways in which those were changing, and above all, they had the data. There was a confidence in the room that well structured, well prepared data which effectively covered the niche in depth was a key strength. By the end of the conference participants were reminding each other that AI would never succeed unless it had access to data of the highest quality. It was all very well to hoover up the dark web. It was apparently okay scrape other peoples websites. It was even okay to pay puny license fees as a gesture to regulatory conformity. But, the players in the room seemed to be saying, we have the real data and we know exactly how important it is. And we also now know that AI is only as good as the quality of the data it ingests allows it to be. AI as a consumer experience will be very popular, and will build very large starter audiences, some of whom will stay as they discover real personal value in its usage. But in business and professional and research marketplaces, things will be very different. Here, in the next 2 to 3 years, there will be real opportunities to lop as much as 30% off processing costs, while adding a similar amount to data value. The title of the Outsell conference was “From hype to reality“. Many members of this community clearly now recognise reality as opportunity.
By the end of the meeting, my mind began to fill with the next projection. How will the use of our AI products and services, as we develop them, change the way in which our end-user clients work? it is clear that major pharma players like BioNtech are rolling out AI research assistants. Similar developments will take place in every industrial and commercial sector. Should we be their partners in these developments? Or will their developments be fuelled by licensing our data and cheating us out of the value to be added to it? Exploring those unresolved problems remains the role of research and thought leadership, but at least, at this point, we are discussing the future with a common agreement that there is a future: in 1985, starting out in strategy advisory work, I faced the initial difficulty of persuading anybody that digital did present any sort of future at all . Even if we acknowledge that we had not yet travelled very far, those of us who left Boston last week seemed sure of the direction of travel and were optimistic about its potential.
Oct
1
AI and scholarly publishing: unfashionable glimpses of hope
Filed Under Uncategorized | Leave a Comment
We were sitting having dinner under the awning of a restaurant in the Place d’Armes, the main square of Luxembourg. Charles Clark, copyright advisor to the Publishers Association was rehearsing the arguments which he and I, as delegates to the European Commission DGXIII Legal Observatory were to make the next day at a meeting in the Batiment Jean Monet. I had just asked what the role of software was going to be in protecting digital copyrights, the waiter had topped up the red wine in our glasses, and the author of Clark on Copyright fell silent for a moment. then the great man declaimed “Maybe in fact the only way that we can regulate a technology is through the use of that technology… The answer to the machine lies in the machine!”
I have always been immensely proud that I was sitting at the table where this edict was first pronounced, at least five years before Charles published a book under this title in 2005. I must confess that I thought it was just a conversational flourish until I heard him use it in the meeting next day, and then in many meetings in the days thereafter. It remains however a notion of real value and power, and it applies much more widely than simply to the notification and protection of copyright, although it remains important there. I had been telling Charles about the ability of my EUROLEX database search software to find keywords in complex text, and the way in which we marked up legal documents with metadata that enabled the software to see how they related to each other and which one had come first. His response is particularly telling at this time, as we are engulfed in a new wave of pessimism about the potentially disruptive effects of a “new” technology. How like the early days of the dotcom boom in the late 1990s are these early years of the understanding of the impact of AI. Unreasonable optimism, stock market hype, lack of political leadership and direction on regulation, media pessimism about the end of all things familiar and a general consensus that this means the end of humans society as we have previously known it on Earth!
Just as the hype is unreasonable, so the pessimism is equally overdone. Perhaps we all need to be more aware of the 50 years or so developmental work which lie behind the current state of what we call artificial intelligence. perhaps we all need to be more aware of the serious homes that could take place, and the developmental track that we need to observe before we get to them. And I think that Giles would say that before we conclude that our own jobs are about to be automated, wilt to look at the problems caused by the machine which can be ameliorated by the machine.
If you are working in scholarly communications, for example, and you do feel pessimistic about the future, then the experience that I had in the middle of last month in Manchester would’ have been a useful tonic. Receiving an award from the UK professional body of scholarly publishing,ALPSP, was a huge and deeply gratifying honour for me personally, but as I looked out over a Conference crowd of 300 people, I also had to reflect upon the dedication and ingenuity in the body of the hall. Later on, Adam Day was honoured for his work at Clear Skies( Papermill Alarm). ( https://clear-skies.co.uk ). As I listened, I thought about an interview which I recently conducted for the Outsell FutureScapes video blog series, when I spoke to Elliot Lumb and Tiago Barros about their work at Research-Signals.com. we know that we have real problems with research integrity: we also know that we have some really clever people developing intelligent solutions. While fresh problems will appear overtime, fresh solutions will as well.
Is anyone under any doubt that we will create fully automated peer review systems which operate more successful than human beings? I have been watching this space since the work of UNSILO in Aarhus almost a decade ago, and I cannot now conceive that we will fail in the search for systems that detect plagiarism, copyright theft or papermill inventions that work at a higher percentage of efficiency than human peer reviewers. While the systems will all require human supervision, audit and checking, they will counter the ability of AI to be misused until we come to a further level of technological development which requires a further wave of watchdog development.
If I am right in this, then surely AI will change the game in every other respect as well. The recent launch by Digital Science of their Papers Pro environment is surely another significant developmental pointer. We are moving pace towards end to end article creation systems. The key question may be a political one: which authority authorises and certifies peer review software on behalf of funders, institutions, and researchers?
In some scientific disciplines, and in some laboratories, the development of the article as a report will become a function of the intelligence in the laboratory network.. In other words, the “article “will be in production from the beginning of the research process and will exist as a series of elements which can be drawn together and updated at will. Then our now elderly attempts at article processing automation – the Scholar One generation – will be replaced by systems which do not just process but actually create most elements of the article. Human intervention in detailing findings and drawing conclusions will of course remain critical: other sections of scientific articles, like methodologies and literature reviews, are already semi or completely automated. Those in the scholarly communications businesses who think that AI is all about data reuse, or, sadly about windfall profits from data sales, have not yet thought through the complete range of potential applications of machine intelligence . Now, surely, is the time, at the dawn of the age of scholarly self publishing, for everyone to think very hard about the role of the journal,
Will the thinking take place which is required across the entire scholarly waterfront in order to find and fund the technologies and the business models which will effectively recast the future for knowledge transfer in our society? Again I find a degree of pessimism that really surprises me. Then I heard from the founders of Scholarly Angels ( https://scholarly-angels.com ) . Seeing experienced entrepreneurs like Andrew Preston, Ben Kaube and Paul Peeters scouting the market for fundable initiatives, and start-ups to incubate is a hugely hopeful sign. Private equity and venture capital will not do this early stage work on its own. And the existing institutions of scholarly communication, when they talk change, too often talk about change as if it is something that happens to everybody else around them, but not to themselves. this may have got them through the “age of digitalisation“: it will not get them through in the age of machine intelligence.
« go back — keep looking »