Enterprise services giant IBM and cloud collaboration and storage provider Box have announced a new partnership, in which IBM will build custom Box Skills using its Watson AI.
Launched in 2017, Box Skills apply context, meaning, and other services to content stored in the Box cloud – for example, object recognition and context for photos, text transcripts for audio files, and facial recognition, topic detection, and transcription services for video.
Under the new partnership, organisations will be able to apply Watson’s cognitive services and intelligence to any content managed within Box’s platform.
Insights from cloud data
At launch, IBM is introducing two new services for building custom Box Skills.
First, custom image insights can analyse image data, “enriching it with classifiers” to make pictures easier to search, while training custom visual models to address specific business needs. In other words, it appears that the system will get smarter the more it is used.
According to Box, an environmental organisation could use this enhanced skill to analyse satellite images of coastal erosion, “quickly detecting areas of most impact, speeding up the time to take action, and reducing the costs of monitoring”.
Meanwhile, custom document insights can automatically tag content stored in Box to help users find information in dense documents, such as research papers, service manuals, and legal papers.
IBM will also build custom solutions that apply Watson via Box’s platform APIs for other use cases. For example, IBM has already built a Watson-powered service that automatically translates documents.
Internet of Business says
The announcement comes in the wake of IBM’s public demonstration of its Project Debater platform, which uses AI to analyse thousands of documents and summarise them into arguments for listeners who lack the time to actively research topics themselves.
In this sense, the Box partnership further lifts the veil on IBM’s cognitive services strategy: its aim is to use Watson and other AI services to take the grind out of intensive research work, opening up insights from the reams of data that exist online and in the cloud, in every form and file type.
In itself, this is a valuable service that may speed up decision-making and help provide organisations with greater intelligence and depth of insight. Indeed, it is a potential boon for every type of organisation in a competitive, data-fuelled age, and both AI and machine learning can help users identify patterns in data that human beings might otherwise overlook.
However, there are risks in effectively outsourcing so much ‘bread and butter’ work to machine intelligence so early in the story of enterprise-scale AI. As we explored in our analysis of Project Debater, turning all such research into a passive, machine-generated or -enhanced process means letting go of human agency at sometimes critical parts of the process.
It also means trusting that machine tools have been properly trained at source to carry out these tasks accurately – such as tag or classify images – and without bias. That said, these systems will get smarter and more accurate with repeated use by customers within their own organisations.
In the long run, turning even human subject experts into passive consumers of machine-sorted information may not always provide the across-the-board benefits that providers envisage, despite AI’s enormous strength in pattern identification.
In many professions – legal services being just one – newly qualified staff are trained on the job by doing exactly this type of ‘donkey work’, teaching them to look for insights themselves, and to understand the minutiae of documents at first hand.
Automating the learning curve, as it were, risks knocking out the first rungs of the career ladder that allow people to climb from trainee to expert. At some point, junior staff will have to analyse documents of every type themselves, but may find they lack the critical skills to do so.
Make no mistake, these are valuable tools and innovations, and discovery can be a slow, thankless, back-breaking, and tedious process that wastes human skills and ingenuity. Automating that and enhancing it with machine intelligence may indeed be transformative for many professions, and help uncover new knowledge.
However, while AI as a discipline has been around since the middle of the 20th Century, we are still at the earliest stages of enterprise deployment at scale in a field that everyone is rushing to join.
So in these early years of the AI-enhanced, automated century, we should be wary of placing too much trust in machine-enhanced data processes too quickly. Unpicking where errors took place in a process of across-the-board automation, enabled by AI-enhanced skills, may be a far more complex and time-consuming process than the original research might have been itself.
At this point in AI’s enterprise history, accepting the output of any automated system passively and uncritically may not be the wisest course of action – valuable and transformative though these new applications may be.