Artificio’s goal in building Seezlo is to create the world’s largest meta-art gallery: a gallery of galleries designed to compile and index every known artwork produced by human hands. We decided to build this for a few reasons. 1) To make art more accessible at a time when many people are not able to access physical art spaces, and many artists experience difficulty finding venues for their art. 2) To make digital and image analytic tools more widely available to all the artists, galleries, museums, curators, collectors, students, researchers and aficionados that make art possible. Perhaps most importantly, though, we at Artificio simply love art! and believe it to be a core component of our common humanity.
The name Seezlo is inspired by -- and pays our homage to -- Peruvian Painter Fernando de Szyszlo, one of the most influential artists of Latin America and the World during the 20th Century.
Seezlo is the first product line of the Artificio Moonshot R&D company.
We are currently exploring this possibility, in addition to adding features for copyright protection and advanced analytics for forgery detection and provenance analysis. As of now however, it is not a Marketplace. The pricing feature is entirely optional.
As a team composed of neuroscientists, AI researchers and artists, we’ve taken a very different approach to the problem of art search & discovery. We’ve developed our own custom algorithms inspired by the computations of the human visual brain, and have rigorously tested our models to see if they “see” the same way as humans. We call this process “brain alignment”, and find that our algorithm does indeed “see” much more like humans than many other modern machine vision systems.
Our Seezlo algorithm is built from scratch, and does not use components from any other model currently available in AI space. In fact, our custom “brain alignment” procedure was designed in many ways to address the limitations of algorithms like CLIP and Google Lens. Seezlo is smaller, sleeker, more robust, and trained with substantially less data than almost any performant algorithm made publicly available to date. Whereas algorithms like CLIP and Google Lens need hundreds of millions, if not billions of training examples (many of which are not available to the public), Seezlo is trained only on the order of a million, publicly available images.
This gives us two competitive advantages: 1) We do not need a gigantic cluster to train our brain-aligned models, thus reducing our carbon footprint; 2) We do not use images made by anyone that has not given us full, transparent, and explicit consent to use their images.
As a team of researchers who all got their start in academia, we are committed to open-source science and development. We are currently working on releasing an open-source version of our code, and would be more than happy to see researchers use our code for many applications, and to stress-test them on their own. In the meantime, we have released a Beta version of an API that is currently available for academics (students), developers and startups.
If you’re interested in this API, please sign up at https://seezlo.artificio.org/signup
We are EXCLUSIVELY opt-in. At Artificio, we consider user privacy and consent crucial to our mission of more human-centric AI. This commitment is not just lip service we pay for public relations, but is built directly into the algorithms that power our services, including Seezlo, which we do not train on any data provided by our users, nor on data whose provenance and copyright we cannot directly establish.
The Generative AI algorithms that we have seen dominate the airwaves in the last year (e.g. DALL-E, GPT) are astounding feats of human engineering, and will no doubt change how we think about the centrality of the human mind in the production of art, and even human creativity more generally.
At Artificio, we believe immensely in the power and potential of Generative AI, but are deeply concerned about some of the ways it is currently being implemented. First and foremost amongst our concerns is the widespread use of training data from unspecified sources that may include copyrighted artworks produced by human creators that have no recourse to challenge the use of their artwork in the training of these algorithms. As a society, we have very few regulatory or technological frameworks that allow us to effectively address this problem, but we believe one way to start is by leveraging more transparent AI to tell us about the inner workings of less transparent AI, and in particular, to leverage more transparent AI to tell us about the kinds of influences that less transparent AI may have learned from.
We designed Seezlo in large part to help connect consumers of art to the human creators that make all art (including generative art) possible. We believe the vast potential of generative AI is even better served by making the very human contributions to AI as clear as possible. While many of the major AI companies currently operating today are more than willing to use the guise of “competition” as an excuse for actively obscuring the parts of their algorithmic design process that rely on human creativity, we at Artificio want to put that creativity front and center. Seezlo, and Artificio’s “brain alignment” technology more generally, is just one step on a road towards a more general “human alignment” process that we hope will make AI more fair, more democratic, and more accessible to the very people whose ingenuity and hard work powers these technologies.
It is currently Free.
You can send us an email to info@artificio.org to say hi or ask us any questions. The best way to stay in touch with our developments is also to Sign-Up to our platform here: https://seezlo.artificio.org/signup
