Tim Salimans

AI Research Scientist

[Grigory Bartosh @GrigoryBartosh]

🚀 Excited to share my @GoogleDeepMind student researcher project: Dual-Rate Diffusion✨

⚡ A simple construction that speeds up both regular diffusion and distilled models by interleaving a heavy context encoder with a light conditional denoiser.

🧵👇

[Tim Salimans @TimSalimans]

@_basjacobs Bij volledige achterwaartse verliesverrekening (liefst inc ander boxen) zou een aanwasbelasting inderdaad efficiënter zijn, maar deze wet straft het nemen van risico (effectief tarief is hoger), is willekeurig (verrekening afhankelijk van timing), en zeer verstorend over de tijd.

[Emiel Hoogeboom @emiel_hoogeboom]

New blog post on our recent paper: Beyond Single Tokens, Distilling Discrete Diffusion.

D-MMD lands on the Pareto frontier of gen PPL vs. diversity, outperforming continuous diffusion distillation approaches — while staying native to discrete tokens.

https://t.co/9exGitXWId

[Tim Salimans @TimSalimans]

@joelbasson No staying in NL

[Tim Salimans @TimSalimans]

Touching down now in SF for onboarding at Anthropic! After 7 great years at Google, I'm excited to take on a new challenge and help make Claude even better. Grateful for everything I learned at Google DeepMind and Brain, looking forward to what's next.

[Tim Salimans @TimSalimans]

Myth: Discrete diffusion models are naturally fast out of the box. 🤔
Reality: Naive sampling is actually quite slow and inefficient. 🐢
Solution: In our new paper, we show how the right step distillation technique unlocks massive speedups. ⚡️👇

[Emiel Hoogeboom @emiel_hoogeboom]

You may think discrete distillation is fundamentally flawed, you are (surprisingly) wrong. 🤯

Meet Discrete Moment Distillation (D-MMD). It is a new method that brings fast, few-step sampling to discrete diffusion models! 🧵👇

[Tim Salimans @TimSalimans]

VAEs are back! 🚀 By co-training a diffusion prior with an encoder and diffusion decoder we obtain a powerful recipe for compressing visual data into a controllable number of bits. By modeling this VAE latent space we obtain SOTA results with smaller models and fewer FLOPs!

[Jonathan Heek @JonathanHeek]

1/6 Introducing Unified Latents: what if your diffusion model's latents were measured in bits? Instead of relying on dimensionality reduction, we learn a latent AE with explicit bitrate control.

Paper: https://t.co/AY0QqRPomI
@emiel_hoogeboom, @TimSalimans

[Tim Salimans @TimSalimans]

@_basjacobs Zou hogere box 3 en schenkbelasting echt "de progressie in het belastingsysteem tegen de minste schade voor de economie bereiken"? Lijkt mij nu al (te) verstorend? Mee eens wat betreft bor,hra,ewf,ozb,erfbelasting.

[Tim Salimans @TimSalimans]

Really awesome to see the full potential of real-time generative modeling being realized! Helping to enable this has been a driving goal of our research in efficient generative models for a long time!

[Google DeepMind @GoogleDeepMind]

What if you could not only watch a generated video, but explore it too? 🌐

Genie 3 is our groundbreaking world model that creates interactive, playable environments from a single text prompt.

From photorealistic landscapes to fantasy realms, the possibilities are endless. 🧵

[Tim Salimans @TimSalimans]

@pedegro Ik zie vooral veel vraagstimulans en weinig waar het aanbod van gaat toenemen. Een gebrek aan geld lijkt mij niet het probleem van onze woningmarkt? Ik had hogere verwachtingen van jullie woningmarktplannen.

[Tim Salimans @TimSalimans]

@erikpjverweij Maakt het hierbij uit dat het "geld binnenstroomt" uit particuliere verhuur ipv een goede baan, onderneming, of erfenis? Mij lijkt de conclusie dat herverdeling via sociale huur nogal inefficiënt is als er per individueel geval een rechtszaak gevoerd dient te worden...

[Tim Salimans @TimSalimans]

@jasonfurman Honest question: if the policy is a (possibly revenue neutral) combination of a VAT and a cut in the corporate income tax, is the effect not similar to a tariff? The VAT-financed tax cut makes the domestically produced car cheaper compared to the imported car.

[Aäron van den Oord @avdnoord]

Veo 2 and the new Imagen 3 update just dropped. Amazing effort by the team!!

[Tim Salimans @TimSalimans]

Are you a current PhD student working in generative modeling? Our team at GDM Amsterdam is looking to hire a student researcher / intern to help us develop the next generation of models. Apply before 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫 𝟏𝟑! https://t.co/gY33wBJwLt

[Tim Salimans @TimSalimans]

On my way to Vancouver for NeurIPS. Looking forward to catching up with everyone and sharing the work we've been doing in making diffusion models fast and efficient!

[Ruiqi Gao @RuiqiGao]

A common question nowadays: Which is better, diffusion or flow matching? 🤔

Our answer: They’re two sides of the same coin. We wrote a blog post to show how diffusion models and Gaussian flow matching are equivalent. That’s great: It means you can use them interchangeably.

[Emiel Hoogeboom @emiel_hoogeboom]

Is pixel diffusion passé?

In 'Simpler Diffusion' (https://t.co/jE0W5qNJaG) , we achieve 1.5 FID on ImageNet512, and SOTA on 128x128 and 256x256.

We ablated out a lot of complexity, making it truly 'simpler'. w/ @tejmensink @JonathanHeek @KayLamerigts @RuiqiGao @TimSalimans

[Sander Dieleman @sedielem]

We are hiring on the Generative Media team in London: https://t.co/rePQxzsGMs

We work on Imagen, Veo, Lyria and all that good stuff. Come work with us! If you're interested, don't delay -- apply before 5PM tomorrow (UK time).

[David Ruhe @djjruhe]

🚀 Interested in time series generation?⏲️Excited to share my @GoogleDeepMind Amsterdam student researcher project: Rolling Diffusion Models!

https://t.co/4UXB428ZYY (to appear at ICML 2024)

Thanks for the great collaboration @emiel_hoogeboom, @JonathanHeek, @TimSalimans! 🧵1/4

[Tim Salimans @TimSalimans]

@FTM_nl Dat het allemaal bij deze eigenaren neerslaat is natuurlijk niet eerlijk, maar dat gemeenten in één klap meer dan 100 miljoen aan waarde kunnen creëren op onze woningmarkt met het doorstrepen van één beperkende regel zegt wel iets over de potentie van verdere liberalisatie, niet?