MIT Media Lab's Anemoia Device: A Scent-Memory Machine
MIT Media Lab has developed a revolutionary prototype, the Anemoia Device, that harnesses the power of generative AI to transform photographs into unique fragrances. This innovative device, dubbed a 'scent-memory machine' by MIT, is a vertical arrangement of three components. Users place an analog photograph in the top section, where the magic begins.
At the heart of the Anemoia Device is an AI-powered computer that analyzes the image with remarkable precision. Users are then guided to craft a personalized prompt using three dials, each offering a unique way to influence the final scent. The bottom section houses a set of pumps connected to fragrance reservoirs, ready to create a custom aroma.
Cyrus Clarke, the MIT researcher behind this invention, explains the concept of distillation that underpins the device. It involves transforming and compressing a dense, layered memory artifact into a scent, capturing the essence of a moment in time. Clarke's interest lies in a specific type of nostalgia called 'anemoia,' which is the longing for a time one has never experienced.
The Anemoia Device has the potential to turn any photographic memory into a scent, but Clarke's focus is on unlived memories. He highlights childhood photographs, family archives, and inherited recipes as ideal sources. The 'anemoia' approach broadens the scope, allowing the system to explore universal themes before delving into more personal and challenging questions.
The machine employs a vision-language model to interpret an image's content, but users have the power to shape the outcome. They start by selecting a viewpoint, whether it's a person or an object. If it's a person, they can specify their age, and if it's an object, they can indicate its lifecycle stage, from 'raw' to 'in-use' or 'decay'. Finally, they assign an emotional tone to the photograph, influencing the scent's character.
During a trial, a participant uploaded a photo of a couple enjoying an apple or pear on stone steps in a garden. The user chose the fruit as the subject, selected 'in use,' and opted for a 'calm' mood. The resulting scent was a delightful blend of spiced apple, pear, and earthy musk, transporting the user to a peaceful autumn setting.
However, Clarke acknowledges that a moment in time is more complex than a machine's literal associations. He emphasizes the importance of a 'shared vocabulary' to guide the system to a meaningful representation before exploring its creative potential. The current prototype includes a scent library with 50 fragrances, ranging from sandalwood and pine forest to old books, leather, and sand, each delivered in one-second increments, offering a wide array of mixtures.
The narrative users build around the photograph can also shape the scent's atmosphere. Clarke illustrates this with an example of two people interpreting the same beach photo differently based on their selections. Clarke's passion for making memories more tangible predates MIT. He previously explored storing memories in plant DNA through his research organization, Grow Your Own Cloud.
He believes that in today's digital age, memories are often externalized and stored in digital infrastructure, accessible but not truly with us. Clarke aims to change this by envisioning two future paths for his prototype. One is a desktop-sized device for at-home memory 'printing,' and the other is an online service for remote photo submission.
Clarke embraces the contradiction of using technology to reconnect us to our senses and nature, challenging the notion that everything must be an attention-stealing machine. He believes in creating new forms of computing that encourage us to pause, breathe, and appreciate the world around us.
