How Google's new AI Agent could help YOUR business

ISSUE 027

Welcome to the 27th edition of Curious Hours

Welcome to the 27th edition of Curious Hours, where every week we dive into the latest developments in artificial intelligence. Just like every other week, this edition is packed with fresh insights and updates from the AI world. Let's dive right in without further delay.

Google DeepMind has introduced an AI agent named SIMA (Scalable Instructable Multiworld Agent) designed for playing video games alongside human players. This revolutionary agent is trained to follow instructions and mimic human gameplay, offering a more natural and engaging gaming experience. SIMA stands out because it can operate in a wide range of 3D virtual environments, handling a variety of tasks from simple movements to complex strategic actions based on natural-language instructions provided by users​​​​.

SIMA was developed in collaboration with eight game studios and trained on nine different video games, including No Man's Sky and Teardown. The agent uses pre-trained vision models and a main model that integrates image-language mapping and video prediction, which allows it to perform instructed tasks by interpreting images on the screen and following user-given, natural-language instructions. SIMA operates without needing access to a game's source code or specialised APIs, relying solely on screen images and language instructions, making it potentially compatible with any virtual environment​​.

One of the key aspects of SIMA is its learning process, which involves behavioural cloning from humans; this means SIMA learns by imitating the behaviour exhibited by human players while adhering to language instructions. The agent has mastered around 600 basic skills including navigation, object interaction, and menu use, and is evaluated on its ability to follow instructions to complete nearly 1500 unique in-game tasks. Despite its current capabilities, the team at DeepMind views SIMA as early-stage research, with the goal to develop more general AI systems that can understand and carry out a wide range of tasks in both online and real-world settings​​.

In the broader context, SIMA is a significant step towards creating more general, capable AI agents that can interact with human users in various virtual environments beyond gaming. Its development reflects ongoing efforts in AI research to build systems that can understand and act upon complex instructions in a manner similar to how humans do, marking a potential shift in how AI can be integrated into everyday activities and entertainment​​.

How companies could use Google’s Deepmind:

AI-Based Product Development:Companies can develop new products or improve existing ones using DeepMind's AI technologies. This could be in various sectors, from gaming and entertainment to finance and logistics. The key is to identify areas where AI can solve complex problems or enhance user experiences.

Collaborative Research and Development: Engaging in collaborative R&D projects with DeepMind or using its technologies could lead to the development of innovative products and solutions that can be commercialised. This could involve applying for research grants, forming partnerships, or working on joint ventures that leverage DeepMind’s AI capabilities.

While direct profit-making from DeepMind as an external entity might not be straightforward since it's a subsidiary of Google focusing on advanced AI research, businesses and individuals can leverage the developments and findings from DeepMind to create AI-driven solutions and services in their respective fields. The key is to stay updated on DeepMind's research, understand its applications, and explore how its AI advancements can be integrated or applied to solve real-world problems or enhance products and services.

We’ve now come to the end of the 27th edition of Curious Hours and if you enjoyed this edition, why not share to friends and family using the link below.

Contact us on email:

[email protected] for any enquires, questions or feedback on our newsletter.

Get in touch personally with our emails:

[email protected] (Founder of Curious Hours and MORAZO AI | Over 300k YouTube Views on AI Meets Football | Author and researcher of Curious Hours Content)

[email protected] (AI Enthusiastic Member & Contributor of Curious Hours)