Nando de Freitas (@NandoDF)

Nando de Freitas @NandoDF

I research intelligence to understand what we are, and to harness it wisely. I work at DeepMind and with CIFAR. Before, I was a UBC and Oxford professor.

Tweets by Nando de Freitas

Hi Murray. Gato is tiny. It needs development. With Lamda, Codex, AlphaCode, AlphaFold, Flamingo, etc, the evidence is strong for the hypothesis that scale is going to take us to AGI, but by scaling I mean addressing the challenges in my tweet, not just more parameters.

Shared by Nando de Freitas at 5/14/2022

Someone’s opinion article. My opinion: It’s all about scale now! The Game is Over! It’s about making these models bigger, safer, compute efficient, faster at sampling, smarter memory, more modalities, INNOVATIVE DATA, on/offline, … 1/N https://t.co/UJxSLZGc71

Shared by Nando de Freitas at 5/14/2022

And thanks to @egrefen @karlmoritz @maxjaderberg Karen Simonyan and Andrew Zisserman, and the leadership at DeepMind and Oxford for believing in scholarships for those in need, being inclusive, and creating a bigger and more diverse pool of talent.

Shared by Nando de Freitas at 2/18/2022

“The Turing Test Is Bad for Business”. This reminds me of when I proposed self-driving cars at a CIFAR meeting in the early 2000s, and Geoff Hinton replied “why don’t people take the bus?” followed by “let’s understand the brain”. Inspiring wisdom. https://t.co/YyzyFRlNSS

Shared by Nando de Freitas at 12/4/2021

Thanks Yuke! You played an important role in the birth of many of these robotics efforts at DeepMind.

Shared by Nando de Freitas at 11/15/2021

The letter is worth reading. No one writes like that on Twitter anymore!

Shared by Nando de Freitas at 11/14/2021

At one point pre-Covid we were going to have AI conferences in Africa. It never happened —- it’s time to ask again: Can we please finally do it? @icmlconf @NeurIPSConf @corl_conf @emnlpmeeting @CVPR @aistats_conf LatAm, India, South East Asia would be nice too

Shared by Nando de Freitas at 11/11/2021

Beautiful visualisation! I would love to see a similar video for the connectivity patterns of pretrained transformers. Also wondering what knowledge would we gain?

Shared by Nando de Freitas at 11/7/2021

An amazing company to work for operating in Africa and America delivering blood, vaccines and medicines with super cool flying robots.

Shared by Nando de Freitas at 11/6/2021

It’s disheartening to see these stats despite all the work many colleagues have put into addressing underrepresentation in AI 😞 It’s so challenging.

Shared by Nando de Freitas at 11/6/2021

Shaking the foundations: delusions in sequence models for interaction and control. I learned so much from Pedro Ortega in this thought-provocative AI project. Great way to spend time with a friend at a London pub. https://t.co/KSUj4Az7t9

Shared by Nando de Freitas at 10/22/2021

Lack of diversity and inclusion continues to be the most pressing issue in machine learning and AI safety. I couldn’t possibly stress this more. Leave it to a single group, and the atrocities will continue to grow.

Shared by Nando de Freitas at 10/20/2021

Amazing ⁦@UniofOxford⁩ !! Oxford changes its start-up policy at last. I’m so happy for the departments, ⁦faculty and students, who I miss dearly. Bravo ⁦@CompSciOxford⁩ - truly number 1 https://t.co/hcPkfjfigR

Shared by Nando de Freitas at 10/17/2021

A thread on the history of convolutional neural nets, neocognitrons, and back-propagation. Both Fukushima and @ylecun played important roles that will be forever recognised. They’re also the nicest, humblest star researchers I’ve had the privilege of meeting.

Shared by Nando de Freitas at 10/16/2021

“Goal: How can we learn a reward function that generalizes across environments and tasks?” I love this data-driven robotics research direction. https://t.co/tj1PFgoVHn

Shared by Nando de Freitas at 9/23/2021

On the continuing quest to understand neural networks, a core AI ingredient, which powers most smart tech, including speech recognition, search, translation, image recognition, and protein folding. https://t.co/9hgJUlWt8W

Shared by Nando de Freitas at 9/13/2021

Fantastic 2nd day at this Stanford Workshop on Foundation Models https://t.co/mZfH6O4Y1y via @YouTube - great talks and panel remarks on some of the most important AI topics of the moment.

Shared by Nando de Freitas at 8/24/2021

Video of yesterday’s Workshop on Foundation Models https://t.co/Pj4Md3z6hg via @YouTube

Shared by Nando de Freitas at 8/24/2021

Timely, informative, good, must read! Info on fewshot learning, amortised inference, autoregressive density estimation, encoder-decoder nets is needed so topic sounds less poorly understood. Great science and engineering got us here, though more is needed. https://t.co/8eUjbEWVuR

Shared by Nando de Freitas at 8/22/2021

Accurate prediction of protein structures and interactions using a three-track neural network https://t.co/nEkWDHZ4Ns

Shared by Nando de Freitas at 7/27/2021

A nice multi-agent reinforcement learning library by ⁦@instadeepai⁩ including our old DIAL method ⁦@iassael⁩ ⁦@j_foerst⁩ ⁦@shimon8282⁩ - built using Acme ⁦ https://t.co/F2XmdcA7WM

Shared by Nando de Freitas at 7/27/2021

On AlphaFold, transformers are amazing! https://t.co/H0r7KZX9gf

Shared by Nando de Freitas at 7/21/2021

Natural language to predict viral mutations that could escape vaccines. https://t.co/AnMfCeQhO6

Shared by Nando de Freitas at 7/10/2021

The future of deep learning, according to its pioneers – Wise insights. https://t.co/uKwKD2ViLn

Shared by Nando de Freitas at 7/6/2021

It’s wonderful to see the research on metalearning and sample efficient adaptive text-to-speech having such a meaningful legacy. @yutianc @iassael @BrendanShilling @scott_e_reed @caglarml @avdnoord @heiga_zen @OriolVinyalsML

Shared by Nando de Freitas at 6/8/2021

Bayesian optimisation played an important role in tuning AlphaGo, spearheaded the field of interactive machine learning, and will continue growing as one of the most important topics in AI because interaction with humans matters.

Shared by Nando de Freitas at 6/8/2021