In an attempt to eliminate misleading representation of AI in ‘stock imagery and pop culture’, Google DeepMind released Visualising AI. By partnering with artists, various artworks depicting AI themes related to research, technologies, and real-world consequences, such as artificial general intelligence (AGI), robotics, neuroscience, sustainability, and generative AI have been created. The images are available on stock-free image sites Unsplash and Pexels. With this, Google DeepMind forays into imagery — a new route that the company is paving for itself. 

Breaking Stereotypes

The portrayal of AI usually overlooks worldwide viewpoints, leading to an absence of diversity that can magnify societal inequalities — visualising AI looks to address this. With a preconceived notion of how AI is depicted through streams of code, blue brains or white robots with men in suits, the repository of new images will contain artist-envisioned images inspired by conversations with scientists, engineers, and ethicists from Google DeepMind. 

All the art work — image and motion graphics — is available for anyone to download free of charge. Since its launch, Visualising AI has engaged with 13 artists and has created over 100 artworks which has gained over 100 million views and 800,000 downloads. These have been used by media outlets, research and civil society organisations. 

The Image Ride 

From robotics, protein folding, and maybe even AGI to this, Google DeepMind’s recent fascination with imagery is quite offbeat. However, Google has experimented with a number of image tools. Google’s text-to-image diffusion model Imagen, which is built on large transformer language models, was released in the beta mode. Google had also launched Muse, a text-to-image transformer model, and StyleDrop which allows the synthesis of images in a specific style using the Muse text-image model. 

Looking at other image generation tools in the market, Google has not been able to have a hold over the market. Midjourney, Stable Diffusion, Dall-E 2, and now with Adobe, Canva and other players, the market is being crowded with consumers spoilt for choice. Furthermore, OpenAI is said to be working on their next image-generation tool which is said to be better than Midjourney. 

While not being an image-generation tool, Visualising AI has entered the global stock images and video market, extending its influence in this domain. In 2022, the market size of worldwide stock images and videos was assessed at approx. $4.96 billion. By 2028, the market is expected to rise to $7.33 billion. The surge in digital media and heightened need for visual content across diverse industries has led to substantial market growth. 

Way Through Partnerships

Recently, Google partnered with Adobe to bring Firefly and Express to Bard. The integration will allow users to generate Firefly images within Bard and these images can be further modified using Express. With Big Techs wanting to diversify into domains that are not their main stream of business, partnerships have been the recent go-to strategy. However, each of the partnerships usually has a larger implication for the company in question. 

OpenAI has been actively partnering with a diverse range of companies, from media to product firms. Notably, they acquired Global Illumination, a product company which employs AI for crafting creative tools. OpenAI’s collaborations extend to news media such as Associated Press and American Journalistic Project too, aiming to gain a foothold in the media industry and enhance model training.

With Visualising AI, Google DeepMind partnering with artists to provide images on Pexels and Unsplash might just be the beginning. While currently it is offering images on AI and other technologies in the same space, it is possible that in the future Google DeepMind might venture into other image categories. Considering how other companies are vigorously pursuing image-generation tools, Google DeepMind might be slowly catching up.