Analysing AI with Frédéric Rose & Faïçal Selka of imki

Released in The Interline’s AI Report 2024, this executive interview with imki is one of a fourteen-part series that sees The Interline quiz executives from companies who have either introduced new AI solutions or added meaningful new AI capabilities into their existing platforms.

For more on artificial intelligence in fashion, download the full AI Report 2023 completely free of charge and ungated.

---

What’s your working definition of AI? Does it differ from the public understanding, which is currently dominated by large language models and generative text-to-image models? And how does that definition manifest itself in your solution(s)?

Faïçal: Artificial intelligence is the theory and development of computer systems trained to simulate human intelligence. The major two types of AI that are driving these changes are generative and predictive AI, both rely on ingesting a large amount of labeled training data. On one hand, predictive AI analyzes the data for correlations and patterns to suggest outcomes and make predictions about future state. On the other hand, generative AI uses a variety of techniques, including deep learning to create new content, such as text, images, or music similar to the trained data. Both brings a range of advantages within business use cases, at imki, we specialize in custom generative AI with the aim of augmenting creative processes, particularly fashion. 

Frédéric: In the vast theater of artificial intelligence, two main players captivate the audience: on one hand, the generalist AI, like a digital Swiss army knife, which ventures through a wide range of creative tasks. MidJourney is the perfect embodiment of this, a pre-teen artist who, with casual ease, juggles styles and forms without ever anchoring itself in a genre.

On the other hand, in contrast, imki creates generative AIs specialized in creative crafts that excel in excellence and focused mastery. Meticulously sculpted, they dedicate their existence to a precise domain, surpassing their generalist counterparts, deepening nuances and refining details with surgical precision. Like a maestro in his niche, each specialized AI controls his art with a virtuosity all its own.

How do you believe fashion can take advantage of this next wave of enterprise technology without compromising on its artistic heritage and cultural legacy? What is it going to look like to train custom-built, fashion-specific AI models that honour the industry’s history of savoir-faire but also tap into new possibilities – both in your domestic market and globally?

Frédéric: At the heart of digital innovation, generative artificial intelligences transform the heritage of luxury brands into modern visions. These systems, true technological guardians, create a link between the past and future possibilities.

These AIs explore brand archives to extract patterns, styles and ideas, often too complex for human analysis. By interpreting this data, they propose designs that respect heritage while adding a touch of modernity. 

Thanks to their ability to analyze and synthesize vast quantities of historical and current data, these AIs manage to enhance patterns and trends that often escape human analysis. In so doing, they generate designs that combine tradition and innovation, subtly incorporating classic elements with resolutely modern concepts. Their expertise enables products to be precisely customized to meet consumer expectations, while respecting the essence of the brand. These digital tools, veritable bridges between past and future, enhance heritage while adapting it to the needs of the present.

Some organisations and individuals – maybe justifiably – are weary of hearing about new technologies that seem to offer a vast possibility space, but then fail to measure up to the hype. Perhaps the best recent examples of that were the pushes into web3 and the metaverse, which did not translate well into value. What parallels, if any, do you see between those and the way that AI is being pitched and positioned today? And what do you see as the quickest way to demonstrate that AI is capable of delivering a real, meaningful return today, as well as in the longer term?

Frédéric: Web 3 and the metaverse are certainly innovative technological advances, rethinking the way we interact with the digital world and augmenting our virtual reality with new dimensions of connection and decentralized ownership. However, these technologies are not in the same revolutionary league as generative AIs.

Generative AIs represent a genuine revolution, not only technological, but also cultural and social. They fundamentally transform content creation, product design and even the personalization of experiences, adapting their production in real time to meet the specific desires and needs of individuals.

Paving the way for a new era of creativity and personalized interaction on a global scale.

If we buy into the transformative vision for AI on the scale Frédéric has just described, within fashion and across society, then it follows that we need to approach it carefully and responsibly – in a way that’s accessible and intelligible for everyone. We’ve already established that we need industry-specific models, so it doesn’t feel like we’ll reach that point by just putting new tools into people’s hands and walking away. What do you see as being the necessary steps to demystify and democratise AI, and to make sure that everyone in fashion (both creators and consumers) can understand its potential, its limits, and where it can best be deployed?

Faïçal: The key points to integrate new technological solutions to operational processes, is to have a good understanding of user needs and a right scope definition of the application. That why it’s advisable to encourage collaboration between AI specialists and fashion professionals. Additionally, educational resources targeted at different fashion roles are crucial. Executives need to understand AI’s potential impact on efficiency, personalization, and cost savings. Designers can benefit from how AI inspires design, analyzes trends to be consumer focus, and optimizes production. At imki we have brought together within the team specialists in AI, machine learning, data science and also fashion specialists such as designer, stylist and garment maker. By fostering collaboration and clear communication, we can empower the fashion industry to leverage AI effectively.

For a lot of readers, “AI” is synonymous with the big, off-the-shelf models – both proprietary and open-source. What are the potential legal and creative risks, to brands, of using something like Stable Diffusion or Midjourney versus your approach? And how do you tackle combining both the artistic and technical components of a brand without creating a different kind of risk: the risk of replacing people rather than augmenting their skills?

Faïçal: On the shelf generative models train on massive datasets, which make them good to spark creativity and can be a great source for designers, but there are potential downsides to be considered and not only limited to legal risk but limited in terms of usage. The specific outputs of these models are hard to predict which can make it difficult to achieve a brand’s specific visual identity and be in balance with trends. At imki, we focus on building custom models trained on data specific to a brand. This mitigates legal risks and allows for much greater control over the creative direction.

AI can handle repetitive tasks, freeing up designer’s time for more strategic and creative work by helping them to visualize concepts. By combining the best of human creativity and AI’s ability to process information and generate variations, brands can achieve a unique visual identity, streamline their design workflow and be more consumer centric. At the end we are providing solutions for brands to augment the creative process, while providing a level of trust and security.

What do you see as the near-term future of AI – both within your solution(s) and in general? Do you believe it will be a transformative class of technologies the way people expect?

Faïçal: Looking forward, I believe in the booming of low/no-code solution allowing user to build custom generative AI models. Users can fine-tune the behavior of generative AI models by providing specific instructions or feeding them relevant data examples. This enables the creation of highly customized AI models tailored to specific needs.

We direct our research effort to create truly original generated AI models. For this purpose, we need to train them with data rich in diversity and context. Unlike tasks with clear solutions, artistic quality is subjective and influenced by popular trends. Capturing these aspects requires high-dimensional datasets that go beyond simple patterns; hence how to select and prepare data for creative applications will be key to future developments at imki.

Frédéric: In the short term, the future of AI looks very promising and likely to transform many sectors. It will continue to be integrated into a variety of fields, improving efficiency, personalization and decision-making. 

As far as imki is concerned, the short-term future of AI looks particularly revolutionary. 

Thanks to imki, AI-assisted design, designers will be able to experiment with innovative shapes, textures and patterns faster and more ecologically. 

By optimizing supply chains and predicting trends, AI could also help minimize waste and improve sustainability in the fashion industry. In short, AI is poised to become an indispensable tool in fashion design, transforming both the creation and consumption of fashion.

Overall, AI is set to be a transformative technology, meeting and exceeding expectations by solving complex problems and opening up new avenues of innovation.