Chief Data Scientist @ bp. Ex-Principal ML engineer @Microsoft. Ph.D in machine learning. @MLStreetTalk pod

Bracknell, England
Joined January 2014
My dream came true and I got to interview my hero in AI, @fchollet! I think we should be speaking more about the limitations of DL and the solutions to discrete and continuous problems.
We spoke with @fchollet about neural program synthesis, the manifold hypothesis, type 1 and type 2 generalisation, the measure of intelligence, the intelligence explosion and Keras! invidio.us/J0p_thJJnoo with @ykilcher @ecsquendor
Show this thread
8
10
1
61
Tim Scarfe retweeted
To celebrate the start of the PROBPROG conference today (probprog.cc), we have put the next iteration of our book on probabilistic programming online: arxiv.org/abs/1809.10756. [1/]
7
167
10
781
Show this thread
🥳New Video🥳Topographic VAEs explicitly structure the latent space into capsules of correlated, topographically organized variables. This allows the model to learn equivariance relations between latent and data spaces. Check out the video💪 invidio.us/pBau7umFhjQ @wellingmax
3
35
5
240
Tim Scarfe retweeted
[new video 🔥] @GoogleAI's "The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning" paper explained! YT: invidio.us/watch?v=0OevSony… Kudos to the authors @hardmaru @yujin_tang for making interactive demos available on their blog!
1
52
3
257
Tim Scarfe retweeted
I had the privilege to appear on @MLStreetTalk special edition on Geometric DL alongside @PetarV_93 @joanbruna and @TacoCohen invidio.us/bIZB1hIJ4u8 Many thanks to wonderful hosts @ecsquendor @ykilcher
1
24
0
160
Show this thread
Tim Scarfe retweeted
It was an honour to appear on the @MLStreetTalk special edition on our Geometric DL blueprint! Super impressed by the production quality -- we even filmed some parts live in Cambridge, despite the ongoing circumstances! Many thanks to @ecsquendor @ykilcher. Hope you like it!
Epic special edition MLST on geometric deep learning! Been in the works since May! with @mmbronstein @PetarV_93 @TacoCohen @joanbruna @ykilcher @ecsquendor invidio.us/bIZB1hIJ4u8
Show this thread
2
21
3
188
The show is finally published! We cover so much ground, and this is a very technical and long show. Use the table of contents on YouTube to navigate about. I am also v. interested in Petar's "Neural Algorithmic Reasoning" work too by the way arxiv.org/abs/2105.02761
Epic special edition MLST on geometric deep learning! Been in the works since May! with @mmbronstein @PetarV_93 @TacoCohen @joanbruna @ykilcher @ecsquendor invidio.us/bIZB1hIJ4u8
Show this thread
4
29
0
110
Tim Scarfe retweeted
We loved listening to and watching this one. Incredibly captivating intro and insightful conversation. Thanks for having Jeff on! @MLStreetTalk
We had a fascinating conversation with Jeff Hawkins from @Numenta discussing his thousand brains theory of intelligence. invidio.us/6VQILbDqaI4 with @DoctorDuggar @NPCollapse
Show this thread
0
8
0
38
Rodney Brooks is brilliant, loved this conversion with Lex. I hope we can get Rodney on MLST one day! @MLStreetTalk
Lex Fridman recently interviewed me (for 2.5 hours) and today published it on youtube. Thanks @lexfridman.
0
2
0
19
Talking with students & others the past few days has brought some clarity to the ways in which the LLMs & associated overpromises suck the oxygen out of the room for all other kinds of research. 1/
6
75
25
314
Show this thread
This one has been months in the making. Hope you enjoy!! 😁
We had a fascinating conversation with Jeff Hawkins from @Numenta discussing his thousand brains theory of intelligence. invidio.us/6VQILbDqaI4 with @DoctorDuggar @NPCollapse
Show this thread
2
0
0
20
Tim Scarfe retweeted
Some tech podcasts I love: For the storytelling: Darknet Diaries w/ @JackRhysider For the friendship & insights: All-In w/ @Jason @chamath @DavidSacks @friedberg For the AI wisdom: ML Street Talk w/ @ykilcher @ecsquendor ++ TWIML AI w/ @samcharrington Robot Brains w/ @pabbeel
70
238
8
2,317
Today's Video: A review of "PonderNet: Learning to Ponder" by @DeepMind. Humans think longer about complex problems than about easy ones, so why not have machines do the same thing? PonderNet learns this end-to-end. invidio.us/nQDZmf2Yb9k @AndreaBanino @BlundellCharles
3
29
1
173
Tim Scarfe retweeted
The paper “Learning Mesh-Based Simulation with Graph Networks” shows how to effectively learn physics simulation on irregular meshes. The model can simulate vastly different systems with significant speedups over the ground truth solver: dpmd.ai/2PubkYS #ICLR2021 (3/3)
18
254
33
1,098
46,519
Show this thread
It took less than a week and it's not even robust to rotations. Even if they improve it, adversarial attacks will always be available. This is what we trade our privacy for. reddit.com/r/MachineLearning…
5
22
1
136
Show this thread
Tim Scarfe retweeted
To do something extraordinary, it's usually enough to put into what you're doing an amount of time and energy that extraordinarily few people would be willing to invest. In fact, that's how most magic tricks work.
12
130
8
794
🔥New Video🔥How Apple scans your phone👀In-depth analysis of iCloud's CSAM detection system, including its weak points. (Hint: it looks surprisingly easy to evade). How much privacy is preserved? Is it worth the trade-off? Watch to find out!💪 invidio.us/z15JLtAuwVI @Apple
3
34
2
151
Tim Scarfe retweeted
"...as fields develop they accumulate huge bodies of knowledge that researchers are meant to be familiar with before they start contributing ...the weight of that knowledge—and the need to keep up with what everyone else is doing—can become crushing." 80000hours.org/podcast/episo…
2
7
0
52