The great Dreamer Danihal Hafner has just announced his retirement from Google, where he worked for nearly a decade.
Prior to his retirement, Mr. Danihal was a staff research scientist at Google DeepMind’s San Francisco office.
His research goal is to “build general-purpose agents that can understand and interact with the world.”
As Google’s world model guru, Danijar led/co-led the development of the Dreamer series (Dreamer, DreamerV3, Dreamer4, etc.).
Danihal Hafner
“Today is my last day at DeepMind,” he wrote in a tweet.
Looking back on his nearly 10 years at Google and DeepMind, Danijar believes “an important chapter has closed.”
In the early days of Google, Danijar primarily participated as a researcher in teams such as Google Research, DeepMind, and the Brain team.
His academic background clearly shows the trajectory of his career development.
Participated in the Brain Team internship since 2016.
Danihal recalls that his first internship was in 2016 with Google’s Brain Team in Mountain View, where he worked with James Davidson and Vincent Vanhoucke.
At the time, there was no viable PPO implementation, so the team debugged and experimented under TensorFlow’s static graph framework.
From 2017 to 2018, Danihal completed a master’s degree at the University of London. During this period, he worked at DeepMind in London.
From 2018 to 2023, Danihal pursued a PhD at the University of Toronto. So he worked for a long time at Brain Team in Toronto.
He co-developed several versions of Dreamer with Mohammad Norouzi, another researcher on the team.
Encounter with transformers
But “I didn’t notice.”
Danihal still remembers the first time Łukasz Kaiser showed them the “big” LSTM-generated Wikipedia page.
ukasz kaiser
Łukasz is a machine learning researcher. He initially worked on the Google Brain team and then joined OpenAI.
In 2018, the Google Brain team published Generating Wikipedia by Summarizing Long Sequences. Łukasz Kaiser is one of the authors and the subject is end-to-end generation of Wikipedia entries.
https://arxiv.org/abs/1801.10198?utm_source=chatgpt.com
Danihal also mentioned an episode of Transformers that he “missed.”
At that time, Ashish Vaswani excitedly introduced the advantages of the new architecture in computational efficiency and suggested applying it to reinforcement learning, but Danijar did not pay attention at that time (I did not try it at that time).
Later, this architecture was named Transformer.
Ashish Vaswani
https://arxiv.org/abs/1706.03762?utm_source=chatgpt.com
Ashish Vaswani is an influential scientist and entrepreneur in the field of AI.
He is the co-founder of Essential AI, but is best known as the co-author of the famous landmark paper “Attending Is All You Need.”
In this paper, we proposed a new, simple network architecture, Transformer, that “completely abandons cyclic and convolutional structures and uses only attention mechanisms,” enabling stronger parallelism and faster training speeds.
The aforementioned Łukasz Kaiser is also one of the lead authors of this landmark paper.
Deep learning research
From 2017 to 2018, Danihal completed a master’s degree at the University of London.
During this time, he worked at DeepMind in London and had the opportunity to participate in research in deep reinforcement learning and generative models, collaborating with Timothy Lillicrap (DeepMind) and Ian Fischer (Google Research).
This experience led us to jointly perfect PlaNet, a model-based reinforcement learning algorithm that can learn a latent world model from pixel observations and perform planning and action selection in the latent space.
timothy lillirap
Danijar then had a long-term interaction with Nicolas Heess and DeepMind’s Adaptive Agents team to systematically organize and implement Karl Friston’s Free Energy Principles (FEP).
nurturing dreamers
Once led by Hinton
From 2018 to 2023, Danihal pursued a PhD at the University of Toronto. During this time, he also conducted research as a jointly trained doctoral student at the University of California, Berkeley.
During his doctoral studies at the University of Toronto, Danijal’s main supervisor was Jimmy Barr, and he was also mentored by Geoffrey Hinton.
During this time, Danihal also worked for a long time at Brain Team in Toronto.
