Why shouldn’t DeepMind be underestimated compared to OpenAI?

Once we understand the value and applicability of AI in both companies, we cannot underestimate DeepMind’s capacity. Over time, Google will discover breakthrough applications, through the powerful Deep Reinforcement Learning capacity that DeepMind owns.

DeepMind’s achievements can be mentioned in a recent study on the application of AI to predict folding proteins – an important issue in the development of new drugs. In healthcare, folding protein is also a perfect field for training AI applications. Specifically, DeepMind’s folding protein prediction system – AlphaFold is trained to use Protein Data Bank data bank, with 3D structure and genetic structure of 150,000 proteins.

In terms of research, both companies are focusing on Deep Reinforcement Learning technology, with the same direction in improving AI. However, it is unfair to compare the technology of the two parties, because deepmind and OpenAI’s algorithmic achievements are equal. Even in the field of gaming, deepmind can be seen to have brought great breakthroughs comparable to the GPT-3 model. However, these works are less well-recognized and recognized by the media.

DeepMind owns more than 1000 employees with hundreds of well-paid Doctors, continuously launching academic research works. Among these works, the most popular is the battle of AI AlphaGo when fighting the world’s top players.


DeepMind can still go further in NLP technology, while creating new language models on a large scale. Up to now, the company is still focusing on improving Google’s language models, and researching and developing more environmental visibility for its AI technologies. These ideas appeared in a new deepmind study called “AlignNet: Unsupervised Entity Alignment”.

It can be said that, with the popularity and high applicableness of GPT-3, OpenAI is considered a leader in this technology war. However, this comparison does not make much sense in AI research rooms in practice, because DeepMind is no less competitive.