Martin Wilson, of Asset Bank has written an article for Information Age: 7 predictions for digital asset management in 2030. The item has some similarities with Toby Martin’s piece which I discussed a couple of months ago (albeit the latter is about the medium term future of DAM well before 2030). I agree with a number of the points Martin mentions (particularly towards the latter half) but not all of them and I think there are some deeper trends which need to be discussed in more detail. This is a fairly in-depth topic, so I shall cover this over a two-part article.
The first observation I will make is that although 2030 sounds like a long time into the future, it is just over 13 years away as I write this in September 2016. A lot of people hold the belief that technology is a fast-moving field and in relative terms, they might be right, but I contend that progress is actually slowing down. I am not alone in holding this view. For those still doubting this thesis, consider that 13 years ago it was late 2003. While there are quite a few differences between now and then, it doesn’t seem substantially different to me and I suspect that someone with average IT skills from 2003 who was transported forward 13 years would not take very long to get up to speed with modern digital technology. The major difference I can think of is that mobile phones have turned into portable computers and their usage has increased as a result. The basic styles of interaction used for many applications, interface conventions, methods of implementing software etc, however, are largely the same.
If you want to further assess the longer-term velocity and trajectory of progress in the tech sector, compare 2003 with 1990: where there were considerably more changes (the biggest of which was mass consumer adoption of the internet). Going back further, in computer technology terms, 1990 and 1977 are virtually unrecognisable from each other. A knowledgeable computer user from 1977 would be completely lost 13 years later. This is a factor which I do not think has been properly considered, not only in this article but by many other commentators too: old tech will generate drag effects which slow the adoption of newer alternatives and these are not being properly factored into assessments of the pace of change. In part this is because there are many more end users than ever before and the tech industry is becoming a victim of its own success. I will return to this point in the second part of this piece, but first, let’s have a look at three of the predictions.
1. DAM interfaces will change dramatically
“Imagine a virtual reality DAM interface that lets you browse through your assets as if you’re in an actual library. Or a holographic interface where you can swipe through the air and make voice commands. DAMs won’t necessarily remain a flat, screen-based experience – by 2030, users could be interacting with their systems in totally new ways.” [Read More]
I can see VR possibly being used but not necessarily as a common method to interact with catalogues of digital assets. Consider how this might work: to interact with assets in a VR environment, you already need to be within it to be a frequent user, otherwise it will involve effort to keep changing from one style of interaction to another. Are hand gestures how you want to be working over an extended period? While there might be physical health reasons for favouring a more active method of operation, if you need to move hundreds of ad-hoc unrelated assets from one collection to another (a task which is quite common for DAM users) then a lot of waving your arms around will be required to do what you can now with some barely perceptible hand movements to control a mouse pointer.
When engaged in these kind of more intellectually demanding exercises, users generally need to refer to documents, open other applications, read emails and consult with colleagues by telephone etc. These sources of information need to be accessible via the same VR interface, otherwise they will be constantly flipping between the real world and the virtual one. With a screen, this is easy, you just look away from it; with a headset, either you need to be able to switch it off or remove the device. Having to do that numerous times every hour will quickly become annoying and I can see many people just deciding to skip the VR and stick to the screen unless there is some obvious benefit on offer.
I think the VR interface will feature (especially for some use-cases) and it might be employed for casual use, in the same way that tablets are for some DAM tasks (where it is helpful to be able to carry them out on the move) but for those who have serious work to do, it will be back to the screen, mouse and keyboard again because it is easier, faster and more reliable. I believe it might take longer than 13 years for this to become widepsread enough for the cost to be justifiable for many DAM users, however, there might be a bigger shift in the search and control interfaces to DAM systems and I will discuss this in part 2.
Similar issues are apparent with voice control also. This method is convenient for some interaction, but not more intensive work. For over sixty years now, an updated version of the typewriter has been used to enter text into computers (even soft keyboards on phones etc have a direct lineage as evidenced by their qwerty key layouts). The reason is obvious: it is simpler, and more accurate than voice control. Crucially, you can also type while having a conversation with someone else other than the computer. I could envisage some hybrid interactions using voice control, however, one example might be ‘macros’ (i.e. common tasks that have been rolled into a single operation) since these would be faster than having to find some deeply hidden menu options which you need to use more frequently. A further factor is if younger users become more familiar with voice control and interfaces are adapted to service their needs as they join the workforce, but I think 13 years is not long enough for a complete change of this nature to fully manifest itself.
On AR, I can foresee Augmented Reality having more of a role to play as this allows you to conduct tasks in the real world simultaneously, although implementing these will not be cheap and whether the benefits are sufficient to convince everyone to give up keyboards and mice remains to be seen. A lot of the issues are similar to VR, although it might offer a compromise or interim option which appeals to some.
It is difficult to escape the conclusion that many technologies like VR, AR and voice control are going to take a long time to lose the ‘gimmick’ tag and are likely to be regarded as an unnecessary indulgence when it comes to tough commercial decisions about whether enterprise budgets should be spent on them. I think they will gradually acquire traction and that process might have started by 2030, but I don’t think they will be fully mainstream by then unless the current innovation velocity trends prove to be temporary. DAM, in particular, is quite a conservative field when it comes to implementation of wider technological innovations. Simply spending any money on the DAM concept at all is something that it has taken years for many users to accept the necessity for. Anything which isn’t paying its rent in terms of ROI on system expenditure, is likely to get set aside until the benefits are comprehensively proven.
2. New types of assets will become prominent
Martin proposes that there will be new types of digital asset that are not conventional nor even exist at present (e.g. the ‘digital scents’ he mentions). I would tend to agree with him on this, it seems reasonable that the scope of digital assets will expand as there is more modelling of the physical environment in binary/digital form. He makes the point about video taking longer to feature in DAM than should have been the case and that DAM developers should read up on these new innovations to prepare for them. With all that said, even after 25 years of DAM technology, images are still the primary asset type held on DAM systems and the most common use case. A lot of the reason for this is based on interfaces and how easy that asset type is to deliver. One of the reasons video adoption was so slow was both the ability to generate preview proxies and the complexity of video formats in general. Video only became practical because Flash was included by default in nearly all browsers and simultaneously bandwidth improved to make it feasible. As such, for new asset types to acquire sufficient traction, they need to be those that are easy to both create and view using cheap commodity technologies.
There probably will be a lot of new asset types, but not all of them may last for a significant duration. Some critical judgement will be required by DAM developers to identify whether they are worth the cost implied by the additional complexity of developing plug-ins to handle them, in addition to issues like whether proxies are necessary (as was the case with video).
3. Artificial intelligence will transform DAM
I have written extensively about AI and the implications for DAM before and I will continue to advance the skeptic case in relation to this subject until I see results which prove it is satisfactory for production use. Between now and 2030, I predict something of an odyssey of discovery for those that wish to utilise automated visual recognition (the duration of which will be roughly similar to the ancient Greek epic also).
“Machine learning will allow DAMs to understand both their assets and their users better. People will no longer have to spend hours uploading, tagging and categorising assets – DAMs will take care of this themselves. ” [Read More]
DAMs won’t entirely ‘take care’ of cataloguing themselves in a way that you can fully trust. Human intelligence will still be required to find and correct mistakes. The AI-generated ones might be harder to identify because some of the output will be correct and some will not. As AI is employed more widely, these limitations with the current tools will become more obvious. DAM vendors will act as first-line support for many of the providers of the underlying AI technology since having promoted their benefits, end users will expect them to get the faults fixed too. That is a point to bare in mind if you plan to use any of them.
There will be a few end users who will gamble that the algorithms work as reliably as the developers of these technologies advertise. Some of them will pass off without incident and be hailed as demonstrations of the fantastic ROI obtainable, others will lead to unforeseen consequences that could have implications for those who depended on them. As discussed in an earlier article about this topic on DAM News, a complex question is who has liability for the faults AI technology will unavoidably throw up.
The way most complex problems get dealt with in other IT systems is to break them down into subsidiary components, which then need to be combined integrated together in a custom manner. This will prove to be more demanding than expected and suggestions from one component will conflict with another, then decisions required about which component’s suggestions to favour and under what circumstances. An approximate comparison is video codecs, there are numerous methods of encoding and decoding videos, formats that are nominally within the same container but differ radically internally and all kinds of other complexity which may require some specialist expertise to deal with. I predict something similar will happen with AI visual recognition tools, but the scale of the problem is quite a lot larger than video.
One of the current AI tool vendors, Clarifai, have begun to tackle this problem by building components and modules for different subjects, e.g. food, weddings etc. Their method is probably the only practical approach, but others will compete with them in the more popular categories especially. Sooner or later (and as happens with DAM) users will want to combine functionality developed by different providers. There are few AI interoperability protocols that have anything close to momentum. Whether any are agreed among AI tool vendors by 2030 remains to be seen, but after 25 years there are still none in DAM and AI has the same mix of small startups and corporate leviathans all operating in the same market.
Well before 2030, end users will be fed up with the effort and expense involved in what they believe to be a ‘simple’ task, so you can expect to hear lines like “It’s only tagging some photos, how hard can it be, why are we having all these problems?”, “Our image search must be broken because I can’t find anything” etc. This is partly because many DAM users underestimate the work involved for human beings to do this task properly, but also because the technology will over-promise and under-deliver for some time to come. By 2030, user expectations might be managed with greater care than is happening now and the results will be improved too, but expect assertions that your DAM will ‘take care’ of cataloguing to be met with some cynicism. As ever with technology, however, at this point those responsible for the AI tools might actually be getting somewhere with it. In this sense, Martin’s prediction might be correct, but I think he has glossed over some of the detail of the period between now and 2030 and discounted negative user reaction to some of the failures and how long this will persist for. In addition, it is far from a foregone conclusion in the manner he has described.