Trending Articles
Trending Topics
Trending
Last night I visited IBM’s rather swish London HQ for an “AI for Business” media briefing. The headline act was Lord Ed Vaizey, who is probably the most digital-savvy Minister for Culture, Communications and Creative Industries that Britain has had. But, as ever with such events, the truly interesting stuff happened around the edges.
On this occasion, it was through demonstrations where IBM showcased its AI for business use.
In this article, I will quickly run through four ways that IBM is using AI in the real world. Some of it is already commercially successful. Some hopefully helping the world in its fight against global warming and human trafficking. Some still at the prototype stage. But all of it is fascinating.
We know IBM works with NASA. Early IBM computers even helped take astronauts to the moon. Now, the two giants are working together to create geospatial foundational models. “Think of ChatGPT,” said Mark Stephens-Row, a Solutions Engineer and Meteorologist at The Weather Company (an IBM company). “It’s kind of like that, but with satellite imagery.”
The difference is that this large language model has been trained on NASA images across multiple decades. “[We] trained that model to recognise things like floods and wildfire scars, and differences between one satellite pass and another over multiple periods of time, for example. It’s petabytes of data.”
Anyone can download the model from Hugging Face. This means that a government or campaigning organisation can use hard data, and modelling, to see the effect of human behaviour on their environment. For instance, the Kenyan government recently proved that protecting an area of rainforest with a fence was having a positive effect on the area.
It isn’t news that IBM has a partnership with Boston Dynamics and owns robot dogs. But rather than weaponise them, IBM loads them with sensors. This means the dogs can patrol inhospitable locations such as nuclear power plants (not, just to be clear, in the reactors!) and monitor for gas leaks, unusual vibrations, temperatures, and capture images to send back to base for analysis.
The role of AI here varies depending on the situation. Some of it is in the control of the robot dogs, so they can make decisions without human intervention. Some of it to do with translating analogue information (such as a reading from an old-fashioned pressure gauge) and converting it into structured data.
But I’ll be honest: the most impressive thing here for me was seeing, yet again, how slickly these robot dogs move.
I had a long chat with Edd Biddle about the work IBM is doing with a Norwegian newspaper, iTromsø. “They’ve got a team of journalists who monitor construction plans in the Tromsø area,” said Edd. But each day, he explained, they’re submerged with information: on one example day, 161 lumps of data that they need to wade through to work out what’s interesting.
IBM’s AI tech not only examines that data for key terms and people, but also ranks it by likely importance and — using generative AI — produces a short summary for the journalists to read. It’s a proof of concept for now, but the time savings for the paper are already clear. It’s likely to move to a commercial deal.
And it’s not only about time savings. “There’s even new stuff that may not have been on their radar,” said Edd.
One of the motivations for this work was to show that generative AI works across languages. IBM’s WatsonX has been trained on 108 different languages, so it hammers home the point that gen AI isn’t just about English.
I also chatted to IBM’s Phil Lewton and Lydia DeFelice from Stop The Traffik, a group that campaigns against human trafficking. Phil is Associate Partner & Garage Leader, Customer Transformation, IBM Consulting, and explained how IBM Watson was helping in the fight.
Similarly to the investigative journalism work, the key here is to turn huge amounts of unstructured data into something humans can act upon. The IBM/Stop The Traffik partnership dates back to 2018, and much of the work behind the scenes is machine learning rather than true AI.
But what matters is that this data can be aggregated and linked — from reports on websites to forum posts to government data — and that data pool can be turned into reports and hotspots. Things people can act upon.
Stop The Traffik is always looking for more data. You can join its Traffik Analysis Hub here.
