Tag Archives: Machine Learning

Never Argue with an AI

I owe our system an apology. After it started forecasting a downturn in Covid-19 cases for Scotland, I was dubious, suggesting that because it didn’t ‘know’ about the combination of the final Euro2020 matches or the relaxing of restrictions at a time of rising case numbers, it was being overly optimistic. I was wrong. I owe it the computational equivalent of a beer or two.

From about 17 June, it was forecasting a peak in case numbers for Scotland on or around 5 July, even if it was short on actual case numbers. Since 22 June though, it’s been spot on in both curve and case numbers. The image is of our case number model, plus our current forecast, overlaid with our forecast from 22 June. Basically, it nailed it. The lesson is not to argue with a system that uses more variables than I can count without taking my socks off.

2021-07-14 Case number projections and hindcast for Scotland
2021-07-14 Case number projections and hindcast for Scotland

 

Conscious Uncoupling

As of 11 May, our Covid-19 forecasting system started consistently predicting a coming rise in cases for the UK, driven largely (as it turns out) by the Delta variant of the Covid-19 virus. In the UK this has almost certainly been facilitated by both the general relaxation in lockdown and by events such as the G7 Summit and the Euro 2020 competition. So far, so bad.

Other factors though are far more hopeful – we are seeing a very significant decoupling between case rates, hospital and ICU admissions and in particular death rates. This does demonstrate that the game has changed, and that we need to be rethinking our approach to managing our way out of the pandemic.

That said, government policy in many places still appears to be conditioned by the assumption that the vaccines effectively eliminate transmission, something that simply has not been demonstrated. Continue reading Conscious Uncoupling

Early Warning Onwards

The UK government stated yesterday (13 May) that rising case numbers in the Bolton area were a cause for concern, and that very many of these cases were of the so-called ‘Indian’ variant (B.1.617.2 being the designated variant of concern, with B.1.617.1 and B.1.617.3 under investigation). Here, raw data for case numbers has been available for weeks, with  organisations such as the Sanger Institute also providing a very informative breakdown by sequenced  variants.

Our analytics platform had identified Bolton and other areas as potential concerns more than two weeks ago and had flagged a correlation between these hot spots and the ethnic balance of the local population, such that, even in the absence of cross-border travel data, of the emerging variants or the situation in India, we were able to provide early warning of emerging problems.

Which very much begs the question as to why the UK government only raised this yesterday, and why travel restrictions from India were only imposed long after the pandemic reached critical mass there. Continue reading Early Warning Onwards

Late to the Party. Again.

SAGE announced today that England’s R number has risen across to between 0.8 and 1. They update their pronouncements once a week, based on their modelling from data that’s even further behind.

We take a different approach: we use emergent and inferential analysis to generate R number calculations and 28-day forecasts, on a daily basis, for every local authority in the UK.

We can say that England, as of today, is at an R number of around 0.92, up from a low of 0.80 on 19 April. Our forecasting suggests that it’s going to go over 1.0 from tomorrow, reaching roughly 1.3 by the end of the month, with England leading the way, followed by Wales and Northern Ireland, with Scotland doing rather better, for the moment at least.

2021-04-23 R Number and forecast for England

Continue reading Late to the Party. Again.

Building a Better Crystal Ball

Another Friday update: we’re well into our private Beta of our predictive analytics and what-if? modelling system for Covid-19 analytics.

So what is it telling us today?

As of 3rd February our projections are (within their confidence limits, which of course become broader the further out we look, even if the central projection is tracking the reality curve well), that the R number bottoms out about now for the UK as a whole, with case numbers continuing to fall until around the 9th, by which time R number is back to .92 and, by the 13th, it’s more likely to be above 1 again, mostly driven by the SE (Essex particularly) and Merseyside (see header picture).

Continue reading Building a Better Crystal Ball

Lagging Decisions, Big Consequences?

On Friday 29th January, the Scottish Government announced that Na h-Eileanan Siar (the Western Isles) is being put into Level 4 lockdown, following a surge of new cases.

On the basis of the data available to us and our modelling approach, we’re not convinced about this decision: it appears to have be made on the basis of out-of-date analysis in an area which turned the corner on this outbreak some time ago.

Our emergent analytics, which generate fresh outlooks every day, suggest that the peak of the outbreak here was passed on 19th January and that it has declined, on multiple metrics, since then.

Continue reading Lagging Decisions, Big Consequences?

Kinetics of a Pandemic

We’ve been thinking for some time about how best to present the dynamic of the pandemic in a way that actually shows what’s happening – the R number doesn’t give any idea of magnitude and is – in our opinion – best kept behind the scenes as a contributor to analytic models, raw or compensated case numbers are just that – daily records – shocking enough in themselves but they still don’t show the energy in the thing. Continue reading Kinetics of a Pandemic

Two Worlds wins research funding for Covid-19 Intelligent Analytics

Two Worlds is one of the successful applicants to a £40M fund created to support “Business-led innovation in response to global disruption”, a competition that attracted 8,600 applicants. Working with a team including epidemiologists, mathematical modelling specialists and the Department of Computer Science at Imperial College, Two Worlds is using udu’s intelligent analytic software to tackle this problem. Continue reading Two Worlds wins research funding for Covid-19 Intelligent Analytics

Wye AI, Man!

Most AI practitioners will argue that the risk to humanity from AI doesn’t (and won’t) come from an AI waking up one day, deciding that the best way to solve the world’s problems is to wipe out humanity and then serendipitously finding that it’s in control of the world’s nuclear weapons. On the principle that cock-up trumps conspiracy, pretty much every time, we’re far more likely to take a range of hits from the misapplication of an AI that’s either too stupid1 to do the job that’s been asked of it or where those deploying it are incapable of understanding its limitations (or indeed don’t care, as long as they’ve cashed out before it all falls apart). Broadly speaking, machine systems fail for one or more of these reasons: Continue reading Wye AI, Man!

Then a Miracle Occurs: The Hype of AI Pitches

I spend quite a lot of my time doing due diligence on innovation funding applications. I’ve been doing this for rather longer than is comfortable to contemplate so, over the years, I’ve seen progressive tides of hype wash in, fill a few rock pools, and then wash out again, only to re-emerge a few years later – assuming it had any merit in the first place – in a form that actually works as part of the overall problem-solving ecosystem. That innovation-development-hype-disappointment cycle may actually happen several times before the rest of the innovations needed for an idea to gain market traction catch up. That’s certainly been the case with VR and AR, with IoT and, most of all, with ArtificiaI Intelligence (AI). Continue reading Then a Miracle Occurs: The Hype of AI Pitches