Science Museum's massive attic clearance finds no magic AI lanterns

Museum group’s relocation of 320,000 items was a tech and logistical success story, but testing times remain

Image:
Science Museum's massive attic clearance finds no magic AI lanterns

Group’s IT lead tells Computing about how the team integrated multiple applications to manage a major relocation of treasured objects, and how difficult economic times mean further challenges ahead.

London's Science Museum is one of the UK's major tourist attractions, with exhibits ranging from Victorian beam engines to the latest space technology. It's run by the Science Museum Group, which operates four other museums too, including the National Railway Museum in York and the Science and Industry Museum in Manchester.

Nearly four million people visited these museums last year, but the exhibits they viewed represent less than 5% of the 7.3 million items in the Group’s collection. The vast majority are held in storage for use by researchers and academia - when they can be located that is.

"We call it the nation's attic," said Tom Saunders, lead software developer at the Science Museum Group.

Recently Saunders and his team played a major part in a daunting multi-year task of shifting 320,000 of these items to a new home in the Science and Innovation Park in Wiltshire, tracking and cataloguing them as they did so.

"Some of these things had been in [the previous building] for a hundred years. We had a record of where we thought things were, we had shelves where things actually were, and they mostly overlapped, but we had to go through take a photo of every object and do a hazard assessment, because plenty of our collection is considered hazardous."

Since the Wi-Fi in the old premises was unreliable, they were unable to load the information directly into the ageing Mimsy object store database, instead having to deploy an interim staging setup. For this they brought in SnapLogic to integrate the various moving parts into logical pipelines. "For two or three years, we were running about 100 different pipelines, just to integrate all of this data," said Saunders.

Image
Description
Tom Saunders, Science Museum Group

Fortunately, their efforts paid off. Having had the cobwebs brushed off, all 320,000 objects are accounted for and safely ensconced in their new home. Another positive upshot of all this hard labour is that there is now a platform for the team to build on.

Many existing enterprise systems and pretty much all new processes are now integrated via SnapLogic, including HR, Payroll, the Shopify ecommerce website, stock and distribution systems, the volunteer management system, data warehouses, ServiceNow ITSM, CRM, and an automated process for integrating photography into the digital asset management system.

The latter came about due a collaboration with Google Arts and Culture, which agreed to digitise The Science Museums' recently acquired photo archive from the Daily Herald.

"They took high-quality digital TIFs of the front of every photo, and also the back, which had handwritten notes to give context. Google is huge and they can scale up in seconds, so they went: 'Yeah, we'll digitise 10,000 images, and then we'll have a press event. Here are 10,000 TIFs, you've got a month till the press event'.

"So we had to scale up the servers on the digital asset management system to be able to cope with ingesting all those files, and SnapLogic just carried on fine. We didn't have to change too much on that."

No magic lamp in the attic

But not everything goes onto the new SnapLogic integration platform. As a non-departmental public body, the Group can little afford unnecessary expenditure. So, if it’s a legacy system and it’s functioning, it's left very much alone.

"We just don't have the resources to go around tearing out things that work," said Saunders.

"Post-Covid, the whole museum sector is suffering for lack of money, and unfortunately we have lost staff. We're also being asked to do more with less - and it's not like no-one asks you to do that when things are going well, too."

Inevitably, this includes "using AI to make money," which, says Saunders, is much easier said than done. "You can't just rub AI on a company and suddenly it becomes better."

The Science Museum Group is already using AI for tasks including transcribing oral histories and filling in gaps in records, and there's huge general interest amongst employees, he went on, but it's hard to know where it will be of real practical use in the museum. Unfortunately the great attic makeover uncovered no magic AI lanterns.

However, with many staff now using GenAI routinely in their private lives, Saunders is hopeful that money-saving use cases will emerge during training courses he's running about its safe use at work.

One idea that shows promise is helping to unify catalogue numbers, whose formats have morphed over the decades, meaning that the same object can be referred to by multiple codes. "It would be a very useful tool if we could just automatically link up object numbers."

At a later stage such a system might be augmented with photos, although the size and diversity of the Group’s collection makes that a big ask.

“We certainly could try object recognition, but we've got, like, 700 scalpels. I'll be very impressed if AI can tell the difference between 700 scalpels from the Victorian era onwards, so I was going to start much simpler than that."