Welcome to Devana.ai, the customizable intelligent assistant designed to increase your knowledge and optimize your ideas. Founded by experts in artificial intelligence and scientific methodology, we're committed to revolutionizing the way information is analyzed and verified.
Explore our site to learn more about our history, our products and how Devana can improve your efficiency and innovation.
Barbara has 3 master's degrees and 1 doctorate. She is a specialist in scientific methodology. Back in 2011, she was already drawing up her revolutionary concept: an intelligent assistant with the power to innovate, capable of reading dense documentation and connected in real time to current events and the web.
In 2017, she founded Scriptor Artis.
In 2019, the company is turning to Deeptech.
It's time to create the intelligent assistant.
Marvin crosses Barbara's path. He's a senior developer. He works at Webedia, in Big Data. As chance would have it, they both specialize in AI: Marvin in Tech, Barbara in Design. Marvin loves the project. They joined forces to boost innovation.
In 2021, Devana was born.
Devana brings together multidisciplinary talents. Each member is imbued with a dual research & innovation culture. This is the key to designing instruments adapted to people, improving their daily lives, activities and well-being.
One spirit, one company. Scriptor Artis, "Giving meaning to information.
Scriptor Artis is first and foremost a concept. This Latin expression means the person who teaches others to transmit the right information. Devana is the fruit of this concept. Each of our members aspires every day to pass on his or her methods and know-how to others.
Innovation, a collective inspiration.
Devana was born of an initial idea, spelled out in the following problem: Can we create an AI that co-pilots the researcher? By researcher, we meant the person who looks for accurate and reliable information, with the aim of producing qualified knowledge. It took over a year of cross-disciplinary research to answer this general question. The mission was twofold, and could be summed up in 2 questions:
1. Is knowledge production a systemic activity?
2. Are GPT-type language models compatible and, above all, promising?
To start work on the LLM (Large Language Model) AI, the mission team obtains OPEN AI's agreement to work on GPT2. We are asked to submit a dossier to prove our development intentions.
From the very first months of testing, work on GPT2 revealed a constraint: If AI masters language, it is capable of hallucinating. By this, we mean that it can simply talk nonsense. While the communication science research team continues its human studies, the AI team begins its first R&D mission: to design a system capable of connecting the AI, an LLM model, in real time to the web and documentation. Indeed, it is imperative that it be contextualized in real time.
In 2020, we bet that AI could accompany us in our knowledge-based activities. But to work, AI needs a model. Our work on knowledge production needs to reveal a systemic activity.
It's all about thinking of intelligence in a plural way, from ontological, psychological, sociological, and computational points of view.
The fusion of intellectual fields between social sciences and formal sciences is necessary to truly understand the issues of artificial intelligence.
Can we model the way we produce information and knowledge?
The researcher's work case is virtuous. There are many well-documented articles on their activities. They deal as much with the standards applied to writing as with the social problems encountered. As a result, the challenge of producing reliable, verified written information calls on a set of parameters, constraints, and contexts that define a common situation. The result is positive.
We can adopt the following sociological reasoning: when a large number of people have different social origins, personalities, educational backgrounds and skills, but encounter the same problems when faced with a similar situation, there is a good chance that these problems are not due to the people themselves but to the situation.
We're looking for an intelligent co-pilot. So it's not just a question of whether our information processing can be modeled, it's also a question of whether AI can support us in this process. After all, our aim is to enhance our capabilities, not replace them.
In 2022, we conducted an in-depth study of the concept of accompaniment.
Accompanying is a process that facilitates learning for the person being accompanied.
Devana must integrate five ideas into its operations:
- The idea of secondarity: The supporter comes after the person being supported. She has no authority.
- The idea of the journey: The supporter moves forward with the person being supported to move towards.
- The idea of complementarity. We do something together.
- The idea of circumstance: the helper helps in a specific circumstance.
- The idea of induction: The coach uses reformulation (generalization) to encourage self-reflexivity.
How can we ensure Devana's successful development? This question echoes a second we're often asked: Is innovating in AI dangerous?
Simply put, an instrument is the tool (artifact) used by a human to perform an activity (pattern). In our case, we're looking to make AI the tool that will enable humans to improve their performance in processing and producing reliable information. The resulting tool is called Devana.
Simply put, an instrument is the tool (artifact) used by a human to perform an activity (pattern). In our case, we're looking to make AI the tool that will enable humans to improve their performance in processing and producing reliable information. The resulting tool is called Devana.
Devana is therefore the mediating instrument between the AI model and the Subject. In our further research, we prefer to speak of an inducer. We also consider the Subject as the individual possessing execution capacities acquired in his social environment. These are called schemas. In an anthropocentric approach, we thus consider the individual as a set of schemas, which will evolve over time, through experiences and activities.
Let's remember that the instrument is the mixed entity that mediates, linking the subject and object, or schema and artifact in an environment. There are 3 dimensions to this mediation: epistemic, pragmatic, and heuristic. In short, the instrument is as much technological as it is psychological.
The most important aspect lies in the exchange between the subject and its artifact. The more exchanges there are, the more the human's activities are likely to evolve (instrumentation), causing the artifact to evolve (instrumentalization) and vice versa. This observation must be integrated into our innovation approach. We speak of a developmental design approach. It's the guarantee of instrument mastery, and hence of innovation in perpetual evolution.
Mastering innovation requires an appropriate scientific approach. Scriptor Artis has chosen the developmental design approach, because the design of an innovation continues in its use.
More concretely, we need to understand that once the tool is in our hands, our ways of doing things change, and as a result, we may use the tool differently, leading to catachresis, i.e. the use of the tool outside its intended purpose.
To summarize, even if the instrument is framed, it's still possible for the user to open a backdoor and make other use of it, hijacking the functions intended by the designers.
Why not understand this phenomenon as an active, participatory contribution by users to their instrument and its use? For us, it's a question of innovating by following the continuous process of instrumental genesis.
So, to master and improve Devana, our innovative instrument, our teams work every day to understand its evolution through its links with users, and to identify new patterns and their influences in a participative design dynamic. More commonly, we refer to this internal approach (and philosophy) as continuous co-design, where innovation becomes a constant and virtuous back-and-forth between the instrument, its users and creators.
