Ivan Lobov

Research Engineer @GoogleDeepMind, ex @Criteo. ML for chip design and chip design for ML

[Ivan Lobov @ilobov]

Overview of our work on logical chip design is out!
See details in a blogpost
https://t.co/L7sNv3fyMB

And in WSJ https://t.co/RVeN1fwhVY

We've been working on it for almost two years and I feel we're just scratching the surface. The domain is enormous and under-explored 1/n

[Ivan Lobov @ilobov]

@y0b1byte Couldn't agree more! That is why our team does just that for more than 3 years now.

[Alexander Novikov @SashaVNovikov]

We are looking for a London-based researcher. During the last couple of years our team did AlphaTensor and FunSearch and our goal hasn't changed: using AI to discover novel algorithms. Join us! https://t.co/K866OwTlc1

[Ivan Lobov @ilobov]

@flavianv Congratulations! 🎉🎉

[Ivan Lobov @ilobov]

We are hiring a Research Engineer to work on AI for Chip Design. Apply if you have strong research/engineer experience on ML and experience/interest on Chip Design.
https://t.co/gPeYfBNDjt

[Jeff Dean @JeffDean]

I’m very excited to share our work on Gemini today! Gemini is a family of multimodal models that demonstrate really strong capabilities across the image, audio, video, and text domains. Our most-capable model, Gemini Ultra, advances the state of the art in 30 of 32 benchmarks, including 10 of 12 popular text and reasoning benchmarks, 9 of 9 image understanding benchmarks, 6 of 6 video understanding benchmarks, and 5 of 5 speech recognition and speech translation benchmarks. Gemini Ultra is the first model to achieve human-expert performance on MMLU across 57 subjects with a score above 90%. It also achieves a new state-of-the-art score of 62.4% on the new MMMU multimodal reasoning benchmark, outperforming the previous best model by more than 5 percentage points.

Gemini was built by an awesome team of people from @GoogleDeepMind, @GoogleResearch, and elsewhere at @Google, and is one of the largest science and engineering efforts we’ve ever undertaken. As one of the two overall technical leads of the Gemini effort, along with my colleague @OriolVinyalsML, I am incredibly proud of the whole team, and we’re so excited to be sharing our work with you today!

There’s quite a lot of different material about Gemini available, starting with:

Main blog post: https://t.co/NzSycJl7aE

60-page technical report authored by th Gemini Team: https://t.co/CEdMRyYSLo

In this thread, I’ll walk you through some of the highlights.

[Ivan Lobov @ilobov]

Our team is looking for a strong research engineer, hardware background is *not* obligatory. Please, share and recommend someone!
https://t.co/AGrF7rwRzZ

[Ivan Lobov @ilobov]

@iris_IGB Энноинг советская ностальгия эсайд, я надеюсь, что это не фейк.

[Google DeepMind @GoogleDeepMind]

Computer chips power our phones, computers and other electronics. ⚡

But producing them can take years of work. 📱

We’re using AI to help design more advanced circuits by treating them like a neural network – and taking their performance to new heights: https://t.co/GRYgp25qE1

[Ivan Lobov @ilobov]

@OlivierTemam And our collaborators: @DJ_Mankowitz, Edouard Leurent, Daniele Calandriello, Michal Valko, @Natashantropova, Paula Kurylowicz, Shariq Iqbal and many others who made this happen! n/n

[Ivan Lobov @ilobov]

Overview of our work on logical chip design is out!
See details in a blogpost
https://t.co/L7sNv3fyMB

And in WSJ https://t.co/RVeN1fwhVY

We've been working on it for almost two years and I feel we're just scratching the surface. The domain is enormous and under-explored 1/n

[Ivan Lobov @ilobov]

Huge shoutout to our core team: Vinod Nair, Georges Rotival, Ngân (NV) Vũ, Marco Gelmi, Kshiteej Mahajan, Adam Hillier, Sergio Guadarama, @OlivierTemam 3/n

[Ivan Lobov @ilobov]

Generative audio models are moving at hypersonic speed this year! 5th of Jan - WALL-E on 60k hours beats quality benchmarks. 8th of Feb - similar performance, but with 15 minutes of audio!!

[ëugene kharitonov 🏴‍☠️ @n0mad_0]

As a result, SPEAR-TTS synthesizes speech with considerably higher speech quality than recent baselines and, in particular, it nearly matches VALL-E in speaker similarity – while having considerably higher speech quality and using 240,000x less parallel data! >>

[ëugene kharitonov 🏴‍☠️ @n0mad_0]

New paper🚨!
SPEAR-TTS is a multi-speaker TTS that can be trained with only 15min of single-speaker parallel data. It synthesises high-quality speech & can generalise zero-shot to a new voice using only a 3s sample. 🧵>>

[Ivan Lobov @ilobov]

Insider wrote about my story, from marketing to ML. Very touched by their interest :)

https://t.co/5izmEIndnN

[xnwup @xnwup]

Today, Hackerone took $25k from me, because I am a belarusian citizen. 1/9

[Ivan Lobov @ilobov]

Oh my god this is beautiful

[chessapig @chessapigbay]

Exciting day today: My first paper is up on Arxiv!! I’m a big boy now

https://t.co/Nx2NjgxRHa

We’re extending a bridge between condensed matter and algebraic geometry, built out of quantum matter on *non-euclidean crystals*

Let me tell you about it, in a friendly 🧵!

1/

[Ivan Lobov @ilobov]

An important reminder that looking stupid is an intrinsic part of learning.
On other hand, I come from quite a diverse background and I often felt like I cannot afford looking stupid. I needed a job no matter what and I adapted the best I could. No regrets https://t.co/GFWAQNGWLc

[Ivan Lobov @ilobov]

I remember I couldn't experiment with Mujoco a while ago because it was behind quite an expensive license. OSS is the best way to democratize science, awesome initiative DM!

[Google DeepMind @GoogleDeepMind]

We’ve acquired the MuJoCo physics simulator (https://t.co/knwXLZMr4L) and are making it free for all, to support research everywhere. MuJoCo is a fast, powerful, easy-to-use, and soon to be open-source simulation tool, designed for robotics research: https://t.co/Of3Q1W2GIR

[Ivan Lobov @ilobov]

@sbos Who will be doing them then? Esp high quality ones.