Mike Walmsley
@mike_walmsley_
Followers
480
Following
10K
Media
42
Statuses
450
Using deep learning to answer astrophysical questions. Dunlap Fellow at Toronto.
Toronto
Joined November 2017
I quit Twitter years ago. I'd like to say it was in protest, but in truth I quit because it made me feel insecure ("look at all the cool research everyone else is doing!") and creates an echo chamber where we all read the same papers. To chat, please email. Good luck all.
0
0
9
AstroML sometimes feels impossible to keep up with. Coming to @NeuripsConf always helps me put that in perspective. None of us are keeping up! That's fine! Dive in the ocean and see what you find!.
0
3
24
The Monkees "I'm a Believer" played at the end of the #NeurIPS23 responsible AI session. Feels optimistic!
0
0
7
RT @Micah_mb010: With awesome collaborators, I'm super excited to share another contribution to the @ML4PhyS workshop at #NeurIPS2023! It's….
0
4
0
2) Finding rare galaxy subclasses by hunting around in Zoobot's embedding space, with @radastrat 📡. I love building useful tools. These explorations help pin down what we ultimately create for @ESA_Euclid and @VRubinObs. @turinginst.@UoMPhysics
1
4
20
Double #NeurIPS2023 #ML4PS arxiv day!. I think the future of @galaxyzoo is to go beyond classification. Here's two steps towards that. 1) Deep learning segmentation of bars and spirals in DESI, with @DrAshleyNova 🦾. @NeurIPSConf.@ML4PhyS
3
10
50
RT @fchollet: It's a symbiotic relationship: AI doomers need AI hype in order to be taken seriously and be part of the discourse, and AI hy….
0
83
0
This Thanksgiving, I'm grateful for all the US computer scientists leaving our GPUs unguarded. A beautiful time of year.
0
0
7
RT @DevinaMohan_: Looking forward to the @ML4PhyS workshop at @NeurIPSConf. In this work we use samples from HMC to examine the cold poster….
0
1
0
RT @MHuertasCompany: I am hiring a postdoc to work on JWST, Euclid and ML at the IAC in Spain. Please, get in touch if interested or let me….
iac.es
0
22
0
Planet 9 was the planet inside us all along 😊.
First-ever: We've identified a new astronomical object, 'Buried Planet', using SEISMOLOGY, rather than telescopes. It's a survivor of Theia, the planet that collided with Earth 4.5 billion years ago to form our Moon. See our @Nature cover paper:
0
0
6
I suspect there's even a Candy Pay Gap. Among my trick-or-treaters, when offered a candy tray, girls were more likely to carefully take exactly one sweet while boys more likely to simply grab handfuls. Feels like a metaphor. Or maybe foreshadowing.
1
0
2
Halloween feels like the perfect setup for econ experiments. Do kids with loads of candy get given less, or more? How about if they have an empty bag? Do groups get less individually? What if there's a parent watching?. All participants get paid in candy!.
1
0
2
It's #ML4PS #NeurIPS2023 notification day and I'm nervous. I have no LLM, ViT, CLIP, diffusion, etc. This time @DrAshleyNova and I made an utterly boring model. It's a bad paper strategy. But it's good for doing science and I'm proud we did it anyway.
2
0
22
Such an elegantly stark illustration.
Historically, much of the gender gap in earnings could be explained by differences in education and occupational choices. However, this year’s economic sciences laureate Claudia Goldin has shown that the bulk of this earnings difference is now between men and women in the same
0
0
1
RT @jwuphysics: Our DESI LOW-Z Survey paper is now published -- and I forgot to write a thread!. In this work, we select very low-z (z<0.03….
0
2
0
In citizen science, and perhaps in society, AI helps people do more. Each @galaxyzoo volunteer click is more valuable than ever. Thank you for making this possible. We've got big plans for new tools that go way beyond classification. Stay tuned.
0
0
5
"Name one astronomy discovery from AI" . I heard this challenge on a recent panel. There's lots of astroML hype, but where's the science results?. This paper is my answer. Our measurements would not have been possible with volunteers alone. The science depends, quietly, on AI.
1
0
4