Magnus Ross's Avatar

Magnus Ross

@magnusar

Postdoc at UCL AI Center. Currently work on: forecasting, uncertainty quantification, antimicrobial resistance. Previously: Gaussian processes, physics informed ML, trading power. https://magnusross.github.io

64
Followers
244
Following
24
Posts
15.11.2024
Joined
Posts Following

Latest posts by Magnus Ross @magnusar

Books I read for fun in 2025 | Magnus Ross

In which I discuss some books I read last year
magnusross.github.io/posts/2025-b...

13.01.2026 11:39 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Preview
Could revealing Rishi Sunak’s cowboy boots ruin Brex-Tex relations? There’s a Sunak in my boots!

It’s a big day for central banks around the world today, so naturally Alphaville is tackling the big topics β€” like whether Rishi Sunak got some cowboy boots as PM and how much they cost. www.ft.com/content/82d2...

18.12.2025 09:23 πŸ‘ 6 πŸ” 4 πŸ’¬ 2 πŸ“Œ 0

🚨 New paper alert 🚨

Our latest work on using machine learning to help clinicians to switch patients from IV to oral antibiotics more efficiently:

openreview.net/pdf?id=j8wsq...

09.12.2025 13:56 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Zero alpha | Magnus Ross

I wrote about a crazy new LLM benchmark where the models are let loose on the crypto markets. Spoiler: sticking all in the prompt does not work that well in this case!

magnusross.github.io/posts/nof1-a...

21.11.2025 09:06 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Moms, Models and Medicine | Magnus Ross A good friend of mine is deep in the world of startups and spends a lot of his time doingΒ idea validationβ€”that is, trying to understand if there is a market for a given idea or product. Despite the fact that in the startup world success is eventually judged by sales or profits, whereas in ML4H it is more likely to be adoption by clinicians and, hopefully, an associated improvement in clinical outcomes, both rely on designing something that people actually want and will use. Therefore, I think many of the tools that help entrepreneurs validate ideas can be repurposed to help researchers undertake projects with real impact.

I wrote something about building systems people actually want as an academic in ML. It's pretty much an open letter to 6-months-ago me.

magnusross.github.io/posts/moms-m...

29.08.2025 13:12 πŸ‘ 2 πŸ” 2 πŸ’¬ 1 πŸ“Œ 1
Preview
The American guns turning Haiti’s gangs into an army Weapons smuggled from Florida are empowering the country’s militias and enabling them to challenge a fragile government

The American guns turning Haiti’s gangs into an army https://on.ft.com/4lVGHFY

30.07.2025 04:03 πŸ‘ 47 πŸ” 28 πŸ’¬ 3 πŸ“Œ 1

The seminar will happen at the UCL AI Centre in the 90 High Holborn building

08.07.2025 14:19 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
A promotional poster for a seminar about foundation models for the Earth system. The image contains a QR code with a link for more details
https://www.eventbrite.co.uk/e/ucl-ai-centre-seminar-wessel-bruinsma-microsoft-research-amsterdam-tickets-1399791095849?aff=oddtdtcreator

A promotional poster for a seminar about foundation models for the Earth system. The image contains a QR code with a link for more details https://www.eventbrite.co.uk/e/ucl-ai-centre-seminar-wessel-bruinsma-microsoft-research-amsterdam-tickets-1399791095849?aff=oddtdtcreator

Come to our seminar with @wessel.ai on "Foundation Models for the Earth System", it's going to be a cracker! It's open to non-UCL people also.

08.07.2025 14:16 πŸ‘ 2 πŸ” 1 πŸ’¬ 1 πŸ“Œ 0

it's the dream blogging crossover!

07.07.2025 16:04 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

Good to get some recognition for my incessant posting of Ben's blogs in the Alphaville comments section!

07.07.2025 15:30 πŸ‘ 2 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Post image

Take note of the legal advice from Amnesty International and more detailed information from @netpol.org below. CND supporters should be aware that point 4 could impact on how supporters refer to non-violent direct action.

07.07.2025 13:15 πŸ‘ 6 πŸ” 8 πŸ’¬ 0 πŸ“Œ 1

great post, please do more like it :)

03.07.2025 16:34 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

ChatGPT is down but The Museum of English Rural Life still stands, proving once again that Silicon Valley cannot compete with the history of rural England and its people.

10.06.2025 12:41 πŸ‘ 13712 πŸ” 2360 πŸ’¬ 89 πŸ“Œ 75

What news editors and politicians think people want: "Just tell them they'll get a tram somewhere in their city if they're nice."

What people really want: "SHOW ME THE FULL GPS ROUTE PLAN COMPLETE WITH LINKS TO THE WIKIPEDIA PAGES OF THE DISUSED RAILWAY LINES THAT COULD COME BACK INTO USE."

04.06.2025 06:25 πŸ‘ 175 πŸ” 13 πŸ’¬ 14 πŸ“Œ 0
Post image

Cancer Research UK today published a report which highlights changes in cancer survival rates over the last 50 years.

Many headlines (like this one from the Guardian) have chosen to report the β€˜doubling of cancer survival rates’ since the 1970s.

But that's misleading...

1/18

03.06.2025 07:21 πŸ‘ 234 πŸ” 113 πŸ’¬ 5 πŸ“Œ 26
Preview
Hardware-Efficient Attention for Fast Decoding LLM decoding is bottlenecked for large batches and long contexts by loading the key-value (KV) cache from high-bandwidth memory, which inflates per-token latency, while the sequential nature of decodi...

A new paper just dropped from Tri Dao(🐐)'s lab!

arxiv.org/abs/2505.21487

Here is my hot take!

30.05.2025 08:04 πŸ‘ 1 πŸ” 1 πŸ’¬ 1 πŸ“Œ 0

the inexplicable core of Starmerism, as a truly new formation in world politics, is seeing governance as wholly an extension of general election campaign messaging

29.05.2025 10:58 πŸ‘ 36 πŸ” 7 πŸ’¬ 7 πŸ“Œ 1
Preview
A foundation model for the Earth system - Nature Aurora, a new large-scale foundation model trained on more than one million hours of diverse geophysical data, outperforms operational forecasts in predicting air quality, ocean wave dynamics, tropica...

Big news!! πŸš€ Aurora, a foundation model for the Earth system, has been published in @nature.com.

A massive congrats to the whole team! Very proud of this achievement, and thrilled to see that it’s finally out there. 😊

www.nature.com/articles/s41...

21.05.2025 16:10 πŸ‘ 26 πŸ” 11 πŸ’¬ 0 πŸ“Œ 1
Preview
The Utter Flimsiness of xAI’s Processes xAI's was happy to put "white genocide" stuff back into Grok's system prompt without second thought.

Short write-up about how I got xAI to briefly add "white genocide" stuff back into Grok's system prompt.
smol.news/p/the-utter-...

16.05.2025 17:36 πŸ‘ 559 πŸ” 135 πŸ’¬ 13 πŸ“Œ 19

it's very odd watching the election news in the UK when Labour still has *years* left on their massive governing majority

they can just ignore this stuff and create prosperity and get popular

and what's particularly weird is that it's not just commentators it's Labour themselves that don't get it

02.05.2025 14:52 πŸ‘ 63 πŸ” 8 πŸ’¬ 4 πŸ“Œ 0
A screenshot of a paper on infectious diseases with the first line of the abstract highlighted. It reads "Like all fields of medicine, Infectious Diseases is rife with dogma that underpins much clinical practice."

A screenshot of a paper on infectious diseases with the first line of the abstract highlighted. It reads "Like all fields of medicine, Infectious Diseases is rife with dogma that underpins much clinical practice."

You know you're in for a banger paper when the this is the first line of the abstract...

28.04.2025 09:17 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Post image

Stealing from X for clout, but someone bought a staggering 25% of the *entire* open interest of a semiconductor ETF (which is 20% NVDA) in a single OTM put option yesterday and today. $3.7 billion of underlying stock, 8ish delta, $23 mil of premium, so levered 15:1 betting on the NVDA embargo news.

16.04.2025 00:07 πŸ‘ 340 πŸ” 78 πŸ’¬ 16 πŸ“Œ 16

one for the debtors!

16.04.2025 10:22 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Forecasting: Principles and Practice, the Pythonic Way

A new Python edition of "Forecasting: Principles and Practice" is now available online at otexts.com/fpppy/. Thanks to @azulgarza.bsky.social, Cristian Challu, Max Mergenthaler, Kin Olivares & Nixtla for making this happen. #forecasting #python

11.04.2025 00:25 πŸ‘ 81 πŸ” 24 πŸ’¬ 3 πŸ“Œ 3

It's frustrating that this stuff isn't picked up by reviewers, and that the Exchange dataset is still being used when it's known to be a flawed benchmark. The fact this model beats the others is irrelevant, because all the models give bad results.

31.03.2025 08:41 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

It makes no sense of the uncertainty to be constant over the forecast horizon here (or in most time series models); the series is extremely unlikely to drop to half of its value in 1 time step, but this model suggests that has a ~5% chance of happening.

31.03.2025 08:41 πŸ‘ 3 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
A grid of plots of time series forecasts, the proposed model's forecasts are highlighted by a red box. The uncertainty estimates for the proposed model are constant over the forecast horizon and are poorly calibrated.

A grid of plots of time series forecasts, the proposed model's forecasts are highlighted by a red box. The uncertainty estimates for the proposed model are constant over the forecast horizon and are poorly calibrated.

I just came across this paper from ICLR 2024 which proposes an intricate combination of transformers and diffusion models to generate forecasts with these uncertainty bounds (red box), which are clearly inappropriate and could likely be outperformed by modelling the data as a random walk...

31.03.2025 08:41 πŸ‘ 6 πŸ” 2 πŸ’¬ 1 πŸ“Œ 0

CUDA mode -> Triton mode

26.03.2025 10:56 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0