Alan Turing Institute to spend UK.gov grants on AI for air traffic control and banking The UK government will bankroll two major research projects led by the country's national institute for AI, the Alan Turing Institute: one to automate air traffic control, and the other for banking services. Project Bluebird, arguably the more interesting of the two, will receive £11.9m and explore how techniques like …
I thought that AI was very hard to show what it was actually doing. As you cannot get much more safety critical than ATC, this will be "interesting"
AI will discover patterns, which might be real or might be side effects of something else that has not been factored into the training data sets.
Context is key. I’d argue you could easily land an A380 on a busy Heathrow runway without railway signalling. Good luck doing it without ATC.
I’m sure the parents of the children involved in the Uberlingen disaster would have something to say about this too.
It's hoped the machine-learning system can assist human controllers in preventing accidents and improving traffic flow. By optimizing journeys, the software may even reduce fuel consumption, too.
Is laudable as long as they don't then use the more efficient flow to increase traffic. Just like fluids, if you have lamimar flow you can pass more, but introduce a bit of turbulence and it all backs up to create a flood.
Ie. They need to make sure they are not creating conditions for a bigger disaster when a trivial incident occurs.
@JassMan - What, you mean use it to demonstrate things like reducing vert/horiz separation, increased flight density (as you suggest) and then adding greater capacity / runways to every airport in the country - all because they can demonstrate that it's "building back better", greener, more fuel efficient and because humanity really, really needs us to be able to spread diseases globally at high capacity. Hmmmm. I mean, what could possibly go wrong? - the latest buzz-words are built into the shiny shiny machines, don'tcha know!. And if and when it does go wrong - it will just be because the silly humins weren't able to react fast enough to thier AI Overloards instructions etc.
Okay, as long as the AI stays entirely in the background unless it raises a potential issue. If it's reporting continuously I predict "bad things" happening. Air Traffic Controllers, as we know, are under immense pressure. They'd essentially be doing their normal job and trying to assess the validity of any AI responses.
Keeping the AI quiet until risk is identified would mean that there's no dispute over the where the decision making responsibilty lies: the human.
I'll admit that at first my blood ran cold, but there are some obvious benefits if done correctly. I reckon the ATC sector is one that might get it right. I'd be magnitudes happier if Turing himself were heading up the projects.
Icon - a toast to Alan Turing (RIP Alan, and thank you for all your unique and wonderful work.)
An interdisciplinary team made up of computer scientists and economists will investigate how AI can revamp financial services and tools, and look for the best ways to apply the technology fairly and securely.
Another fork would be their invention and maintenance of novel and noble and Nobel worthy services, or would that be too controversial and subversive because ....... IT Offers AI B00Sters in Futured Futures Markets. Where everything Heavenly is Provided for Quantum Leaping Progress. A BillionairesRUs Investment Vehicle for Group Hosting and Sub-Contracted Hire …..for Virtually Live Application Engagement of AI B00Sters. ‽ .
A Killer Application that, whether Primarily an Exotic Erotic Eastern Confection or Wild Wacky Western Delight, or not, if AIDeliverable in Private Pirate Promotion and/or as a More Extraordinary Renegade Rogue Rendition.
That's certainly where some are presently at with exercises currently quietly active and stealthily engaged to run riot in such fields. I Kid U Not.
For what?
I'm guessing fraud detection & prevention and maximising profits (when to trade most profitably). Not entirely altruistic.
"Using real data is risky, and isn’t possible for privacy reasons."
I'd suggest "isn't possible" is a bit strong, but fair go, the data isn't staying in-house (two parties) and you're erring on the side of caution rather than relying on anonymising real data. You couldn't have a chat with uk.gov for us could you, about a little project they're running atm.
You know, I don't really want to entrust my ATC or finances to some dumb schmuck's idea of an AI they can get away with foisting on me to give them a career. I'd much rather see AI developed first for things like sorting and grading recycled rubbish, giving my garden mower the ability to distinguish weeds from flowers, that sort of less life-changing shit where there is less to lose when the inevitable titsup happens.
...if you get AI to study outcomes and the ways they were achieved by human ATC bods, then all you've ended up with is a system that will 'think' like a good controller given the constraints of performance and sep'n standards, and in the future, come up with solutions that the ATCO at the time will probably have already (or shortly thereafter) come up with anyway. All you've added is a distraction.
Any further improvement needs to see a change of separation standards, and/or the building of lots more runways.
OT: Last century, the Oz ATC PHBs in 'Bullsh1t Castle' decided that there were parallels between trains and aircraft. They got in a train expert and asked him how to improve efficiency in moving aircraft. After an appropriate period, his (because they were always 'hims' in those days) report said that aviation had it all wrong. Railways had (generally) two lines between main stations, and lots of platforms at each end. Aviation had lots of airways and only one or two runways at each end, and hence had it all backwards.
This post has been deleted by its author
AI: "FU891 Ascend to FL 140 and turn left 160"
Pilot (having a bad day): "Unable"
AI: "I'm sorry, I didn't quite understand that"
Pilot: "UNABLE"
AI: "I'm sorry, I didn't quite understand that"
Pilot: "MAYDAY MAYDAY MAYDAY. We have lost power in both engines"
AI: "If you require a human controller, please say Yes"
....
<i>The UK government will, to the tune of £60m, bankroll two major research projects ...</i>
If these really are MAJOR research projects, from which conclusions would be drawn that change the way we run air traffic control and banking, then £60m is chump change compared to what would be needed for a meaningful in-depth analysis of:
a) how they work (this is AI remember)
b) whether they work
c) how we check that they work and against what criteria
d) what could be the side effects: expected and unexpected
e) how we back out if it's a disaster
f) the costs at every stage: design; implementation; running; disaster recovery
This is just two minutes off the top of my head. £60m is meaningless in this context.
Ah yes. A wonderfully hopeful aim, laudable even. Unfortunately big Gov plus public money plus software plus people ensures the continuation of the usual outcome. Big waste of time and effort followed by a report from the NAO saying the same.
I promise to say I told you so, assuming I lived through the experiment
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
