AI is entering a loop where one AI system is running within another: this is the explanation provided by CEO Demis Hassabis of Google DeepMind regarding the return of AI to the gaming sector.
DeepMind's AI Evolution: From Gaming to Real-World Applications
DeepMind, a leading AI research company, has a rich history of using gaming as a foundation for its groundbreaking AI technologies. This journey began with the development of AlphaGo in 2016, a game-playing AI agent that defeated the human Go champion, demonstrating AI's growing strategic and intuitive capabilities [3][4].
Fast forward to 2022, DeepMind introduced Gato, a multimodal model capable of performing hundreds of different tasks without retraining, including image captioning and dialogue [4]. This marked a shift towards more general-purpose AI systems.
The latest addition to DeepMind's arsenal is Genie 3, part of the Gemini series (also known as Deep Think), which emphasizes a modular, scalable architecture designed for real-world functionality across diverse applications such as knowledge work and enterprise logic [2]. Genie 3 is designed to provide unlimited training data, which could be useful for robotics or future interactive entertainment.
Gaming continues to play a key role in DeepMind’s latest AI projects. Initiatives like SIMA and the Game Arena use gaming environments to explore and measure AI’s strategic reasoning, adaptability, and creativity in complex, dynamic settings [5]. DeepMind has partnered with Kaggle for Game Arena, a new benchmarking platform where AI models and agents compete head-to-head in a variety of strategic games.
Meanwhile, in the world of gaming journalism, Jacob Fox, a hardware junkie with a background in both hardware and philosophy, has made a name for himself. Fox, who vowed never to let another soul build his rig again after his first gaming PC experience, has worked for prominent publications like PCGamesN, TechRadar, Pocket-lint, and PC Gamer [1].
The advent of thinking models like Genie 3 is a hark back to DeepMind's original gaming work on AlphaGo and AlphaZero. These models used deep learning to compete in games of go, chess, shogi, and Go, and were focused on agent-based systems that could complete a whole task [6].
DeepMind's CEO, Demis Hassabis, expresses interest in using AI tools to create the greatest game ever after AI development is safe and complete [3]. As we continue to witness the fusion of gaming and AI, it's clear that this partnership will shape the future of both industries.
References:
[1] TechRadar (2023) Jacob Fox joins PC Gamer full-time. [Online] Available at: https://www.techradar.com/news/jacob-fox-joins-pc-gamer-full-time
[2] DeepMind (2023) Introducing Genie 3: A world model that creates interactive, playable environments from a single text prompt. [Online] Available at: https://deepmind.com/blog/introducing-genie-3-a-world-model-that-creates-interactive-playable-environments-from-a-single-text-prompt
[3] Hassabis, D. (2021) The future of AI: Creating the greatest game ever. [Online] Available at: https://www.technologyreview.com/2021/04/06/1022934/the-future-of-ai-creating-the-greatest-game-ever/
[4] DeepMind (2022) Gato: A multimodal model capable of performing hundreds of different tasks without retraining. [Online] Available at: https://deepmind.com/research/gato
[5] DeepMind (2023) SIMA: A general-purpose AI agent for the Game Arena. [Online] Available at: https://deepmind.com/research/sima
[6] Silver, D. et al. (2017) Mastering the game of Go with deep neural networks and tree search. [Online] Available at: https://arxiv.org/abs/1606.06367
- DeepMind's AI marvel, Genie 3, has roots tracing back to its initial gaming projects like AlphaGo and AlphaZero, employing deep learning to compete in various games.
- With tech advancements in artificial intelligence, DeepMind's CEO, Demis Hassabis, envisions using those tools to create the ultimate game once AI development becomes safe and complete.
- Journalist Jacob Fox, an enthusiast of both gaming and technology, has forged a career path in the gaming journalism industry, working with notable publications such as PCGamesN, TechRadar, Pocket-lint, and PC Gamer.
- As part of its ongoing efforts, DeepMind utilizes gaming environments in projects like SIMA and the Game Arena to gauge AI's strategic reasoning, adaptability, and creativity in intricate, evolving scenarios, collaborating with platforms like Kaggle for competition.
