jörn jacobsen @jh_jacobsen Twitter profile

Last Seen Profiles

@zero_perks

@louismandylor

@delideff

@calistacharms

@thebrogmyster

@fumani218289

@ANT1TV

@min9yx_k

@draxper

@LindaVerweij3

@UndetectableAI

@MfonBas03680305

@smartprakas

@Cz7Ye

@rayanlawfirm

@Nxpxt77

@InformaticaUNLP

@HToutk74774

@GretchenOO8

@Sheriff_RMFC

@jandakembangstw

@brosephtimus

@vigiIantechris

@handofayesha

@Eta

@TTHChilumizine

@YorgosKaramihos

@jpnewspr

@SCATT3RDR0SE

@xww9wwx

@ragdush

@SaaSaa756227723

@rSoftwareGore

@Danicaslaps28

@tomaarsen

@helpme_thebigt

jörn jacobsen

@jh_jacobsen

4 years

A growing body of work focuses on striking differences between current ML models and biological intelligence. We review the literature and argue that many of the most iconic failures can be understood as a consequence of the same underlying principle: “shortcut learning”

6

163

613

jörn jacobsen

@jh_jacobsen

6 years

1/6 Deep classifiers seem to be extremely invariant to *task-relevant* changes. We can change the content of any ImageNet image, without changing model predictions over the 1000 classes at all. Blog post @ . with @JensBehrmann Rich Zemel @MatthiasBethge

Deep Classifiers Ignore Almost Everything They See (and how we may be able to fix it)

Jörn Jacobsen, Jens Behrmann, Rich Zemel, and Matthias Bethge — 25.3.2019

medium.com

4

189

524

jörn jacobsen

@jh_jacobsen

10 months

Our team is hiring LLM researchers and engineers. You'll have lots of opportunities for impact *and* will be able to publish! Opening says Seattle, but location is flexible. Zürich is an option, too. Feel free to DM for questions!

8

36

276

jörn jacobsen

@jh_jacobsen

2 years

Considering to move into AI/ML research but have a non-AI/ML background? Check out the Apple AI/ML residency program (application deadline 7th of Dec):

4

55

252

jörn jacobsen

@jh_jacobsen

2 years

We are hiring an ML Research Engineer to work on the forefront of AI in the health space. If you are passionate about building robust DL models, want to work with outstanding ML researchers + engineers, want to have real impact with your work: apply!

6

55

249

jörn jacobsen

@jh_jacobsen

3 years

If you are considering to move into AI/ML research but have a non-AI/ML background, check out the Apple AI/ML residency program (application deadline 15th of Dec):

4

73

230

jörn jacobsen

@jh_jacobsen

1 year

Official posting is out! We have multiple machine learning research internship positions in our team at Apple Zürich in Switzerland. Please apply here within the next 2 weeks if you would like to be considered:

10

47

227

jörn jacobsen

@jh_jacobsen

5 years

Long-awaited and beautiful paper on "Invariant Risk Minimization" by Arjovsky et al. studies relationship between invariance, causality and the many pitfalls of ERM when biasing models to simple functions. Love the Socratic dialogue the paper ends with...

0

56

209

jörn jacobsen

@jh_jacobsen

4 years

@andrew_n_carr surprise: google will reject your application even if you have a PhdD, are a highly cited researcher, bring in ideal qualification for your job but can't solve leetcode problems.

2

0

187

jörn jacobsen

@jh_jacobsen

5 years

We introduce Residual Flows, an approach based on invertible ResNets that is competitive with state-of-the-art flow models and dramatically increases efficiency over vanilla iResNets. With @rtqichen , @JensBehrmann , @DavidDuvenaud Paper:

1

36

157

jörn jacobsen

@jh_jacobsen

4 years

Invertible Neural Nets (INNs) / Normalizing Flows are amazing! But are INNs always invertible? Surprisingly we find that they often violate this constraint! Below a Glow reconstruction on CelebA with @JensBehrmann @PaulVicol @kcjacksonwang @RogerGrosse

4

35

147

jörn jacobsen

@jh_jacobsen

5 years

Core ML/AI is oversaturated. If I'd look for PhD positions now, I'd look for ML-heavy positions in less populated adjacent fields. E.g. opportunities in natural sciences. It's often a good idea to work on something not everyone is working on already. Don't be a 🐏, be unique!

hardmaru

@hardmaru

5 years

“To start a PhD in ML, without insider referral, you need to do work equivalent to half of a PhD. Hence, in Apr 2019, I decided to dedicate all my time until Jan 2020 to publish in either NeurIPS or ICLR. If I fail, I would become a JavaScript programmer.” — @andreas_madsen ‼️

14

138

768

5

10

135

jörn jacobsen

@jh_jacobsen

5 years

We release code for Invertible Residual Networks: If you are at #ICML2019 check out long oral Wed 2:00-2:20PM @ Hall A and poster Wed 6:30-9:00PM @ PB #11 to discuss @icmlconf ! Joint work with @JensBehrmann , @wgrathwohl , @rtqichen and @DavidDuvenaud

1

23

110

jörn jacobsen

@jh_jacobsen

4 years

Interest in domain/out-of-distribution generalization and algorithmic fairness has skyrocketed over the last years, but with relatively little overlap. We focus on exchanging lessons between the sub-fields and show they can be mutually beneficial.

3

10

82

jörn jacobsen

@jh_jacobsen

1 year

If you are excited about Synergies between Scientific and Machine Learning Models () and looking for internships in beautiful Zürich starting asap - feel free to DM for opportunities in our team at @Apple Zürich!

4

15

86

jörn jacobsen

@jh_jacobsen

4 years

Neat ICML paper ending the expressivity discussion for iResNets and neuralODEs tldr: both are universal approximators for homeomorphisms when embedding inputs into 2*d-dimensions and for non-invertible functions when adding a linear layer on top of that

Approximation Capabilities of Neural ODEs and Invertible Residual Networks

Neural ODEs and i-ResNet are recently proposed methods for enforcing invertibility of residual neural models. Having a generic technique for constructing invertible models can open new avenues for...

arxiv.org

4

11

77

jörn jacobsen

@jh_jacobsen

5 years

Very cool talk by David Duvenaud on the stories around and behind Neural ODEs, wish I attended that workshop.

0

18

74

jörn jacobsen

@jh_jacobsen

4 years

Exciting 40 author (👀) @Google paper on the trouble with underspecification in ML. Providing further intriguing empirical support for many of the issues we raised in our shortcut learning paper () and more. Highly recommended read:

1

13

74

jörn jacobsen

@jh_jacobsen

3 years

Sometimes it's worth refining and resubmitting work. After some bitter rejections, REx finally made it into @icmlconf 2021 as long presentation!! 💫 Kudos to @DavidSKrueger for driving this

David Krueger

@DavidSKrueger

5 years

Want to generalize Out-of-Distribution? Try INCREASING risk(=loss) on training environments with higher performance! Joint work with: @ethancaballero @jh_jacobsen @yayitsamyzhang @jjbinas @LPRmi @AaronCourville

2

34

156

4

6

72

jörn jacobsen

@jh_jacobsen

5 years

"This is the first rigorous proof of identifiability in the context of VAEs ... The advantage of the new framework over typical deep latent-variable models used with VAEs is that we actually recover the original latents, thus providing principled "disentanglement"." 😮

Brundage Bot

@BrundageBot

5 years

Variational Autoencoders and Nonlinear ICA: A Unifying Framework. Ilyes Khemakhem, Diederik P. Kingma, and Aapo Hyvärinen

1

3

14

0

3

64

jörn jacobsen

@jh_jacobsen

5 years

Check out the updated paper and code of Residual Flows for invertible generative modeling! Release includes SOTA-level pre-trained models for MNIST/CIFAR10/Imagenet/CelebA-HQ 🔥

Ricky T. Q. Chen

@RickyTQChen

5 years

We're releasing code *and pretrained models* for Residual Flows, a SOTA invertible generative model, at . Compared to existing flow models that enforce structured Jacobians, we can use simple ResNets and efficient estimators to get unbiased log-densities.

2

46

222

0

13

58

jörn jacobsen

@jh_jacobsen

4 years

Last two years were the best years of my research life so far, thanks to the amazing community at @VectorInst (Toronto is a beautiful place as well) I highly encourage everyone to apply!! Feel free to reach out to me if you have any questions about being a Vector postdoc

Amir-massoud Farahmand

@SoloGen

4 years

We have several postdoc positions at the Vector Institute. If you are a rising star in #MachineLearning , we want you to be here! The deadline for this round is June 12th. After this, we have another round in September/October.

5

27

79

0

2

55

jörn jacobsen

@jh_jacobsen

5 years

1/5 New work w/ @EthanFetaya and Rich Zemel suggests likelihood-based conditional generative models will not solve robust classification. We show competitive models can be easily fooled, revealing fundamental issues with their learned representations and the likelihood objective.

1

7

54

jörn jacobsen

@jh_jacobsen

4 years

Work with Robert Geirhos, @clmich , Rich Zemel, @wielandbr , @MatthiasBethge and Felix Wichmann. Paper here:

Shortcut Learning in Deep Neural Networks

Deep learning has triggered the current rise of artificial intelligence and is the workhorse of today's machine intelligence. Numerous success stories have rapidly spread all over science,...

arxiv.org

5

6

47

jörn jacobsen

@jh_jacobsen

4 years

Very happy that our work on exploding inverses in invertible neural networks has finally been accepted to #AISTATS2021 !!

jörn jacobsen

@jh_jacobsen

4 years

Invertible Neural Nets (INNs) / Normalizing Flows are amazing! But are INNs always invertible? Surprisingly we find that they often violate this constraint! Below a Glow reconstruction on CelebA with @JensBehrmann @PaulVicol @kcjacksonwang @RogerGrosse

4

35

147

0

2

50

jörn jacobsen

@jh_jacobsen

4 years

Following threads on bias in ML I'm surprised how controversial simple facts can be. Makes me glad we wrote this piece on how almost every part of pipeline may contribute to it Many people do, but I wish *everyone* would appreciate how hard of a problem this is with no easy fix!

jörn jacobsen

@jh_jacobsen

4 years

A growing body of work focuses on striking differences between current ML models and biological intelligence. We review the literature and argue that many of the most iconic failures can be understood as a consequence of the same underlying principle: “shortcut learning”

6

163

613

0

7

42

jörn jacobsen

@jh_jacobsen

5 years

👉 @bethgelab showed we don't need to beat baselines to get papers published and make meaningful contributions to science. Negative results, especially if accompanied by new benchmarks, are just as important as positive results. Nicely summarized here:

0

7

46

jörn jacobsen

@jh_jacobsen

6 years

6/6 If you are interested in the details, check out the full paper. It is going to be presented @iclr2019 : Much work to be done to better understand the role of excessive invariance for generalization and adversarial vulnerability!

2

5

43

jörn jacobsen

@jh_jacobsen

4 years

Shortcuts are decision rules that perform well on standard benchmarks but fail to transfer to more challenging testing conditions. Shortcut opportunities come in many flavours and are ubiquitous across datasets and application domains

2

5

45

jörn jacobsen

@jh_jacobsen

3 years

Thought provoking read comparing gDRO/JTT/rebalancing: "these data balancing baselines achieve state-of-the-art-accuracy, while being faster to train and requiring no additional hyper-parameters" > We have actually observed this as well on many problems

Simple data balancing achieves competitive worst-group-accuracy

We study the problem of learning classifiers that perform well across (known or unknown) groups of data. After observing that common worst-group-accuracy datasets suffer from substantial...

arxiv.org

1

5

44

jörn jacobsen

@jh_jacobsen

4 years

These 15 minutes are such a nicely done summary of why connecting densities with anomaly detection is typically most likely not a good idea:

Charline Le Lan, Laurent Dinh · Perfect density models cannot guarantee anomaly detection

slideslive.com

2

7

43

jörn jacobsen

@jh_jacobsen

4 years

Shortcuts arise from model bias + underspecified solution space and manifest themselves as misalignment of intended and learned solution. Interestingly this is not unique to ML but common in biological systems. We discuss connections to comp. psychology, education and linguistics

1

5

41

jörn jacobsen

@jh_jacobsen

1 year

Lots of AI research focuses on massive un-curated data - what about highly curated data? Scientific models are just that: large amounts of carefully curated and summarized experimental data. But how can ML leverage them effectively? Join our ICML workshop:

2

8

38

jörn jacobsen

@jh_jacobsen

6 years

Nice summary by @wielandbr of his BagNets ICLR19 paper. To learn robust classifiers that do not only rely on local statistics, we may need to consider stronger inductive biases and most importantly move beyond improving plain classification accuracy.

0

8

36

jörn jacobsen

@jh_jacobsen

6 years

3/6 We have stumbled upon what may be the first analytical adversarial attack. Our approach allows to arbitrarily change image content without changing the logit outputs at all. Middle row shows images with logits of top-row, but content of bottom-row.

1

9

33

jörn jacobsen

@jh_jacobsen

5 years

Looking forward to @iclr2019 in NOLA and discussing: a) Exploiting Excessive Invariance Caused by Norm-bounded Adversarial Robustness Monday: b) Excessive Invariance Causes Adversarial Vulnerability Tuesday:

Exploiting Excessive Invariance caused by Norm-Bounded Adversarial...

Adversarial examples are malicious inputs crafted to cause a model to misclassify them. Their most common instantiation, "perturbation-based" adversarial examples introduce changes to the input...

arxiv.org

1

2

31

jörn jacobsen

@jh_jacobsen

4 years

Neat paper obtaining invariance guarantees from data augmentation (image translations and rotations, audio volume change ...) by generalizing randomized smoothing to structured transformations:

1

5

29

jörn jacobsen

@jh_jacobsen

5 years

Fantastic summary of the state of the art and open problems in normalizing flows!

cs.LG Papers

@arxiv_cs_LG

5 years

Normalizing Flows for Probabilistic Modeling and Inference. George Papamakarios, Eric Nalisnick, Danilo Jimenez Rezende, Shakir Mohamed, and Balaji Lakshminarayanan

1

12

27

0

4

28

jörn jacobsen

@jh_jacobsen

5 years

Impressive how far Lp-norm robustness can take deep nets. Turns out it's a pretty useful inductive bias, making gradients more aligned with our (semantic) expectations, e.g. enabling the use of discriminative models for image editing and other quasi-generative tasks.

Aleksander Madry

@aleks_madry

5 years

Robustness goes beyond security: Representations induced by robust networks can align much more closely with human perception, and enable simple feature visualization and manipulation. See: (w/ @logan_engstrom @andrew_ilyas @ShibaniSan @tsiprasd B. Tran)

0

49

181

0

3

29

jörn jacobsen

@jh_jacobsen

6 years

Super excited to be a speaker alongside a stellar lineup at the ICML "Workshop on Invertible Neural Nets and Normalizing Flows". Check out the call for papers and consider submitting, deadline is April 26:

0

2

30

jörn jacobsen

@jh_jacobsen

4 years

@geoffreyhinton and an exclusive deal with Azure computing

0

25

jörn jacobsen

@jh_jacobsen

2 years

@bneyshabur Replies seem largely biased towards asking "do I need PhD to land tech-job / have impact" answer is probably: no. In my opinion big strength of non-toxic PhD environments is to foster open-ended research while paying your rent. Doing a PhD can be rewarding experience on its own

1

27

jörn jacobsen

@jh_jacobsen

4 years

Awesome panel @iclr_conf Very interesting discussion on how progress in robustness, fairness and privacy are not simply algorithmic challenges and the need to think about the systems algorithms are embedded in. Cool to see such topics taking center stage!

1

3

27

jörn jacobsen

@jh_jacobsen

6 years

1/4 Norm-bounded robustness can cause invariance-based vulnerability. We are able to find adversarial examples within robust epsilon balls around data! Paper: with: @JensBehrmann , Nicholas Carlini, Florian Tramèr, @NicolasPapernot

1

6

25

jörn jacobsen

@jh_jacobsen

6 years

@karpathy Another very interesting paper from the same lab shows how reducing this texture bias can significantly increase robustness and accuracy: . It's an oral at ICLR19.

ImageNet-trained CNNs are biased towards texture; increasing shape...

ImageNet-trained CNNs are biased towards object texture (instead of shape like humans). Overcoming this major difference between human and machine vision yields improved detection performance and...

openreview.net

1

3

27

jörn jacobsen

@jh_jacobsen

4 years

Take home messages: 1) Analytical invertibility does not necessarily imply numerical invertibility 2) Different tasks have different requirements on invertibility (e.g. local vs. global) 3) Controlling stability is crucial for principled and successful application of INNs

1

2

26

jörn jacobsen

@jh_jacobsen

6 years

4/6 We call the phenomenon invariance-based adversarial examples, a complementary viewpoint to the classical perturbation-based case. We ask: which task-relevant directions is my classifier invariant to? Instead of: which task-irrelevant directions is my classifier sensitive to?

1

4

23

jörn jacobsen

@jh_jacobsen

5 years

This fantastic work is more evidence that only looking at either "robust" or "unrobust" features exclusively is suboptimal. Wouldn't it be great to have models which are able to consider *all* predictive features in their decision 🤖

Aleksander Madry

@aleks_madry

5 years

What if adversarial examples are not bugs...but features? Read about our new perspective and surprising experiments: [1/3] (joint work with @andrew_ilyas @ShibaniSan @tsiprasd @logan_engstrom Brandon Tran)

2

36

109

1

3

21

jörn jacobsen

@jh_jacobsen

5 years

Cool work treating augmented normalizing flows as latent variable models. Full of interesting insights!

Brundage Bot

@BrundageBot

5 years

Augmented Normalizing Flows: Bridging the Gap Between Generative Flows and Latent Variable Models. Chin-Wei Huang, Laurent Dinh, and Aaron Courville

1

0

10

0

1

22

jörn jacobsen

@jh_jacobsen

6 years

5/6 An information-theoretic analysis reveals that cross-entropy is (in part) responsible for this, as it does not discourage such invariance. We extend the objective with an independence term to explicitly allow to control invariance. This fixes the problem in various settings.

1

4

20

jörn jacobsen

@jh_jacobsen

4 years

Very impressive!! Still lots of domain knowledge in data augmentations though, because (as conclusion acknowledges): "With great flexibility comes great overfitting" 💡

0

1

21

jörn jacobsen

@jh_jacobsen

4 years

R2: "the flaws outweigh the merits" mentions not a single flaw, only merits in review

1

0

19

jörn jacobsen

@jh_jacobsen

4 years

All this and much more in our new work: "Understanding and Mitigating Exploding Inverses in Invertible Neural Networks" Link: 👩‍🔬 We hope our work encourages researchers to consider stability as an important ingredient of INN design 👨‍🔬

Understanding and Mitigating Exploding Inverses in Invertible...

Invertible neural networks (INNs) have been used to design generative models, implement memory-saving gradient computation, and solve inverse problems. In this work, we show that commonly-used INN...

arxiv.org

2

20

jörn jacobsen

@jh_jacobsen

5 years

Looking forward to spending the next days @NeurIPSConf . We will present recent work on Residual Flows, Lipschitz constrained convolutional networks and (non-)invertibility of invertible neural networks:

1

21

jörn jacobsen

@jh_jacobsen

4 years

🤯

OpenAI

@OpenAI

4 years

Introducing Jukebox, a neural net that generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles. We're releasing a tool for everyone to explore the generated samples, as well as the model and code:

219

2K

8K

1

3

18

jörn jacobsen

@jh_jacobsen

1 year

It’s a mistake to think what people shaping the mainstream narrative are doing now is what you should aspire to as well - you will always be behind that way. Don’t play their game, come up with a new one.

0

1

18

jörn jacobsen

@jh_jacobsen

2 years

If you are interested in health and its intersection with ML, have experience in biomedical engineering, with sensing hardware, or other fields mentioned in the link, we might even work together (co-mentored by @heinzedeml )! Feel free to DM with questions

0

6

17

jörn jacobsen

@jh_jacobsen

5 years

Takeaway: we need more principled approaches for selecting meaningful robustness bounds and for measuring progress towards more robust models. Awesome collaboration with @florian_tramer , @JensBehrmann , Nicholas Carlini, @NicolasPapernot - pre-print 👉

Fundamental Tradeoffs between Invariance and Sensitivity to...

Adversarial examples are malicious inputs crafted to induce misclassification. Commonly studied sensitivity-based adversarial examples introduce semantically-small changes to an input that result...

arxiv.org

0

4

17

jörn jacobsen

@jh_jacobsen

5 years

@qiyang_li @CemAnil1 @james_r_lucas @RogerGrosse "On the Invertibility of Invertible Neural Networks" ML with Guarantees workshop (Saturday, West Ballroom B) with @JensBehrmann , @PaulVicol , @kcjacksonwang and @RogerGrosse

1

16

jörn jacobsen

@jh_jacobsen

5 years

We prove a fundamental tradeoff between invariance and sensitivity to p-norm perturbations akin to the example above. p-norm / oracle misalignment means there will always exist adversarial examples either sensitivity- or invariance-based, no matter how robust the model

1

13

jörn jacobsen

@jh_jacobsen

4 years

Based on our observations, we develop a set of recommendations for model interpretation and benchmarking and highlight recent advances in ML to improve robustness and transferability from the lab to real-world applications

1

2

14

jörn jacobsen

@jh_jacobsen

4 years

Pleased to share that our paper "Shortcut learning in deep neural networks" has been published as a @nature Machine Intelligence Perspective: PDF access without paywall:

0

1

15

jörn jacobsen

@jh_jacobsen

3 years

A pinch of slightly obfuscated domain knowledge is all you need? 😏

Patches Are All You Need?

Although convolutional networks have been the dominant architecture for vision tasks for many years, recent experiments have shown that Transformer-based models, most notably the Vision Transformer...

openreview.net

2

15

jörn jacobsen

@jh_jacobsen

4 years

We hope our work serves as stepping stone for connecting the dots between seemingly disparate failure modes of current ML models, motivates more research in this direction and justifies why we need strong generalization tests as part of our standard model evaluation protocol

1

2

15

jörn jacobsen

@jh_jacobsen

5 years

"Preventing Gradient Attenuation in Lipschitz Constrained Convolutional Networks" Poster #149 (Thu 10:45am, East Exh. Hall B+C) by awesome *undergrads* @qiyang_li and Saminul Haque w/ @CemAnil1 , @james_r_lucas , @RogerGrosse

1

3

14

jörn jacobsen

@jh_jacobsen

4 years

Derivatives of inverses can become arbitrarily large => "exploding inverse" This can lead to analytical invertibility not carrying through to the numerics, INNs become non-invertible! We explain this effect by analysing bi-Lipschitz properties of common invertible networks

1

2

14

jörn jacobsen

@jh_jacobsen

5 years

Underfull \hbox (badness 4779)

1

0

14

jörn jacobsen

@jh_jacobsen

5 years

"Residual Flows for Invertible Generative Modeling", Spotlight (Tue 4:40, West Exh. Hall C) and Poster #85 (Tue 5:30, East exh. Hall B+C) presented by @rtqichen work w/ @JensBehrmann and @DavidDuvenaud

1

14

jörn jacobsen

@jh_jacobsen

5 years

We also show that increased robustness to epsilon perturbations leads models to ignore important features. We alter images semantically *within* norm-balls and show "robust" models fail on these invariance-attacks while undefended and less robust models do much better

1

0

11

jörn jacobsen

@jh_jacobsen

4 years

We also find striking differences between INNs. Additive coupling blocks stably train with memory-saving gradients, while affine couplings lead to incorrect gradient computation, highlighting the importance to understand influence of architectural choices on exploding inverses

1

12

jörn jacobsen

@jh_jacobsen

4 years

For NFs we often want density estimates on samples not from the training data => We need global invertibility! Indeed NFs can suffer from exploding inverses on OOD inputs implying meaningless density estimates. Solving this requires stable architectures like Residual Flows!

1

12

jörn jacobsen

@jh_jacobsen

4 years

ICML reviews have been updated in CMT. @icmlconf great idea, but why no email about this so everyone knows it happened?

0

4

12

jörn jacobsen

@jh_jacobsen

2 years

decent programme name

1

0

12

jörn jacobsen

@jh_jacobsen

6 years

2/6 To show this, we design an invertible classifier with a simplified read-out structure. This allows us to combine logits (Zs here) of one image with everything the classifier does not look at (Zn here) from another image, invert and inspect the result.

1

3

11

jörn jacobsen

@jh_jacobsen

1 year

What Kyle says :) don’t miss out on SynS & ML! ICML 2023 Workshop, July 28 2023, Room 320 of the Hawai‘i Convention Center #ICML2023 @icmlconf

Kyle Cranmer

@KyleCranmer

1 year

I’m experiencing some major FOMO for not going to #ICML . I was originally planning to participate in the excellent “Synergy of Scientific and Machine Learning Modeling” Workshop (which you should attend) but needed to scale back

1

27

0

2

11

jörn jacobsen

@jh_jacobsen

2 years

tired: back of the envelope calculations of SSL-hours required for tiny humans to reach AGI wired: acknowledging intelligence and invention as inherently social phenomena

0

1

11

jörn jacobsen

@jh_jacobsen

10 months

Here separately the two openings Research: Engineering:

1

2

11

jörn jacobsen

@jh_jacobsen

2 years

Location is flexible: we are open to hire in Cupertino, Seattle, NYC, Zürich. If you have any questions, my DMs are open.

3

1

10

jörn jacobsen

@jh_jacobsen

5 years

Increasing expressiveness of iResNets increases bias of the density estimate. Our main contribution is to introduce an unbiased estimator for the infinite sum in the log density evaluation of residual blocks, alleviating the need for trading off bias and expressiveness.

1

9

jörn jacobsen

@jh_jacobsen

4 years

@poolio Invariant Risk Minimization

0

10

jörn jacobsen

@jh_jacobsen

4 years

"Out of 1800 candidate sequences from the GPT-2 language model, we extracted over 600 that were memorized from the public training data ... Many of these examples are memorized even though they appear infrequently in the training dataset" - N. Carlini

Privacy Considerations in Large Language Models

Posted by Nicholas Carlini, Research Scientist, Google Research Machine learning-based language models trained to predict the next word in a senten...

research.google

1

2

8

jörn jacobsen

@jh_jacobsen

3 years

Candidates from many disciplines are invited to apply (non-exhaustive list can be found under link). If your background is in physics, computational modelling, applied math or electrical engineering, there is a chance that we might even work together :)

0

9

jörn jacobsen

@jh_jacobsen

4 years

Looking forward to using my upcoming days off to maximally take advantage of the new "hard" lockdown

1

0

8

jörn jacobsen

@jh_jacobsen

4 years

Because memory-saving backprop only requires accurate invertibility on training data, we propose an architecture-agnostic solution ensuring local invertibility: bi-directional finite differences penalties But this is not enough for Normalizing Flows (NFs)!

1

9

jörn jacobsen

@jh_jacobsen

5 years

Presented as a contributed talk at the #ICML2019 Workshop on Invertible Neural Nets and Normalizing Flows @icmlconf :

0

1

8

jörn jacobsen

@jh_jacobsen

1 year

We have a fantastic line-up of speakers and panelists: @ln2cool , @_amiller_ , @FrankNoeBerlin , @laurezanna , @vdbergrianne , @sercanarik , @KyleCranmer , Patrick Gallinari

1

0

8

jörn jacobsen

@jh_jacobsen

4 years

Our piece on pitfalls of attributing expert-level radiologist intelligence to pigeons, how failures are part of intelligent problem solving, requiring deep analysis & how progress should start with Q: Should a task be solved in the first place and if so, should it be done w AI?

The Gradient

@gradientpub

4 years

Neural Nets can often succeed on datasets, while failing to actually do the intended task. How? In our latest piece, @jh_jacobsen , Robert Geirhos, and @clmich expand on the concept of "Shortcuts" as a unifying way of thinking about such failures:

3

48

134

1

4

7

jörn jacobsen

@jh_jacobsen

2 years

cc @QueerinAI @AiDisability @black_in_ai @Khipu_AI @DeepIndaba @_LXAI @WiMLworkshop @women_in_ai

0

1

7

jörn jacobsen

@jh_jacobsen

4 years

Code here:

GitHub - asteroidhouse/INN-exploding-inverses: Code for Understanding and Mitigating Exploding...

Code for Understanding and Mitigating Exploding Inverses in Invertible Neural Networks (AISTATS 2021) http://arxiv.org/abs/2006.09347 - asteroidhouse/INN-exploding-inverses

github.com

0

7

jörn jacobsen

@jh_jacobsen

1 year

Great to see adversarial examples having their long overdue comeback

0

7

jörn jacobsen

@jh_jacobsen

2 years

🔥 work showing how to use generative models for music composition in a way that doesn't feel superficial - after all, these models are still tools that require lots of talent to create meaningful art. @patttten seems to have a lot of that

patten - In Me

'𝐌𝐢𝐫𝐚𝐠𝐞 𝐅𝐌' - the 1st album fully made from text-to-audio AI samples. Out April 14th via 555-5555. Single 2, ‘In Me’ & self-directed AI-assisted vide...

www.youtube.com

1

2

6

jörn jacobsen

@jh_jacobsen

4 years

Finally @jmgilmer arguing for more diverse test sets and making call to adversarial robust optimization crowd to focus on real-world distribution shifts. @timnitGebru adding we need to incentivise dataset creation more if we want access to such diverse datasets

1

0

6

jörn jacobsen

@jh_jacobsen

10 months

Thought provoking and important read! Really enjoyed this paper

Hailey Joren

@HaileyJoren

10 months

Models are trained on costly data and require this data at prediction time. We should be able to opt-out and understand the gains of opting in! In our latest w @nagpalchirag @kat_heller @berkustun we introduce models that give users this informed consent #NeurIPS2023 Spotlight

1

33

1

8

jörn jacobsen

@jh_jacobsen

5 years

We also discuss some desiderata for Lipschitz activations and found a normalized Swish nonlinearity to work very well. Additionally, we generalize iResNets and spectral normalization to induced mixed norms and allow the p-norm orders to be learned along with the model.

1

0

5

jörn jacobsen

@jh_jacobsen

5 years

@tdietterich @EdwardDixon3 @icmlconf @CShorten30 @alexisfink @alexott_en Exactly the point we are making with our info-theory analysis and shiftMNIST dataset! All this shouldn't come as surprise when considering the discriminative objectives we are using. However, we do believe and show there are ways to overcome this without modelling whole of p(x).

0

6

jörn jacobsen

@jh_jacobsen

1 year

And a great organising committee: @WehenkelAntoine , Emily Fox, @anujkarpatne , Victoriya Kashtanova, Sharon Di, @EBezenac , @naoya_takeishi , @glouppe

1

6

jörn jacobsen

@jh_jacobsen

4 years

@pfau Also had this discussion recently with some co-authors and ended up agreeing that this is an area where Nature MI Reviews/Perspective articles might be a decent choice to fill this gap in the publication landscape.

1

0

4

jörn jacobsen

@jh_jacobsen

1 year

Topic areas broadly include: * Causal representation learning * Fair and robust machine learning * Combining scientific models with machine learning * Physics-informed machine learning * Structured generative models

1

0

5

jörn jacobsen

@jh_jacobsen

5 years

5/5 Finally, we discuss how multiple of these undesirable properties are direct consequences of the likelihood objective. We conclude that likelihood may be fundamentally at odds with robust generalization in conditional generative models. Paper here:

Understanding the Limitations of Conditional Generative Models

Class-conditional generative models hold promise to overcome the shortcomings of their discriminative counterparts. They are a natural choice to solve discriminative tasks in a robust manner as...

arxiv.org

0

5