dynamic FEATure sensing in Self-Educating Neural Signal Entities

Información del proyecto

FEATSENSE

Identificador del acuerdo de subvención: 101034891

DOI

10.3030/101034891

Proyecto cerrado

Fecha de la firma de la CE 25 Febrero 2021

Fecha de inicio 1 Febrero 2021

Fecha de finalización 31 Julio 2022

Financiado con arreglo a

• EXCELLENT SCIENCE - Future and Emerging Technologies (FET)

Coste total

Sin datos

Aportación de la UE

€ 100 000,00

Coordinado por

IntuiCell AB
Sweden

Periodic Reporting for period 1 - FEATSENSE (dynamic FEATure sensing in Self-Educating Neural Signal Entities)

Período documentado: 2021-02-01 hasta 2022-07-31

Resumen del contexto y de los objetivos generales del proyecto

FEATSENSE represents a disruptive innovation in the field of AI. With novel results on the mechanisms by which the brain predicts the world, IntuiCell (a spinoff from LU) is developing a next-generation AI architecture that effectively avoids well-known issues in the machine learning domain. As IntuiCell dives deeper into within FEATSENSE and beyond, the design and underlying mechanisms of AI are very different from how the brain works.

Problem: Existing AI architectures hinders deployment of intelligent, broad applications and limits SOTA AI to isolated tasks. Specifically, modern AI face serious issues when interacting with the real-world. Despite remarkable engineering achievements, modern AI boils down to statistical models that are goal oriented and rely on pretrained experiences. By discarding all information viewed as unnecessary for the isolated task, AI systems operate under a rigid set of constraints and limitations. This has massive implications in real-world situations - as the real world presents endless unforeseen situations.

Societal benefits: FEATSENSE aims for a reinvention of the field of AI. Specifically, IntuiCell aims to lead the path from task-specific skill to general intelligence, enabling intelligent applications with huge economic and societal impacts. The proposed novel AI architecture has a very high transformative potential, not only for new applications but also for improving performance of existing applications. Thus, the proposed AI architecture constitutes a major technological breakthrough.

Overall objectives: FEATSENSE allows IntuiCell to take the initial and very important steps from a major research endeavour into an early-stage deeptech company, starting the process that leading to a commercial innovation
- Build a business and market strategy for IntuiCell AB.
- Create working benchmarks and example solutions, and
- Target key partners to establish a presence in the machine intelligence/AI industry

Trabajo realizado desde el comienzo del proyecto hasta el final del período abarcado por el informe y los principales resultados hasta la fecha

- Initial ethics report: Outlining possible ethical issues related to development and future deployment of IntuiCell's technology.
- Hire a CEO: Recruitment of Viktor Luthman, having 12 years’ experience from building startups in multiple industries.
- Market report: Outlining major challenges for existing AI technology, why deep neural networks (DNNs) remain dominant in the field of AI/machine learning despite their weaknesses, and possible opportunities for IntuiCell's unique technology to go above and beyond contemporary AI.
- Dissemination and exploitation plan / Business plan: Outlines IntuiCell's initial thoughts on business and strategy, including long term vision and goals, current operations including strategy to grow the team, market and competition, business model, IP strategy, marketing and communication and financial planning.
- Benchmark report: Initial benchmark between IntuiCell's biology-inspired feature extractor that acts as an alternative solution to SOTA (Mel-based techniques), with benchmark comparisons for text-dependent speaker verification. Also explores the performance of these feature extractors under dynamic and static noise conditions.
- Demo application: Presents IntuiCell's first demo application within the first input modality (audio data). This is an end-user oriented, real-time demonstrator of a speakerID use-case for banking.
- Website: IntuiCell's website was built and designed to attract interest from all potential stakeholders, while not disclosing sensitive information about the technology.
- Customer introduction material: In addition to the updated website, IntuiCell has created promotion material in the form of several pitch decks to introduce the company to various stakeholders, ranging from potential investors to future key partners.
- Customer status report: Outlining the potential key partners that has been identified and approached during FEATSENSE, summarising the strategic discussions held with the intent to establish a market and customer base for IntuiCell forward

Avances que van más allá del estado de la técnica e impacto potencial esperado (incluida la repercusión socioeconómica y las implicaciones sociales más amplias del proyecto hasta la fecha)

FEATSENSE mainly concerns business development efforts, aiming to set the strategy and direction for the company on the path toward commercial innovation.
In parallel to the efforts that lie within the scope of the project, IntuiCell's tech development team is working diligently to advance the fundamental technology, with crucial steps taken during the course of FEATSENSE.
Based on results from FET projects ('ph-coding' project, coordinated by prof Henrik Jörntell at LU; as well as FEATSENSE) and other efforts outside the scope of FET projects, we have made important progress on IntuiCell's core tech, indicating very promising opportunities for real-time applications.

At the end of the project, IntuiCell has been successful to demonstrate a first industry use-case within SpeakerID, as well as to source a highly relevant Deeptech industry pilot to demonstrate the competitive claims / advantages of IntuiCell architecture compared to modern AI technology.

IntuiCell holds transformative potential for the broader industry and society, developing novel technology that allows us to take a big step from the current machine learning-era and bringing us closer to a biologically realistic implementation of artificial general intelligence, AGI. IntuiCell presents a fundamentally new approach to artificial intelligence, efforts that are crucial to ensure explainable, fair, ethical, and efficient AI. The technical development activities in FEATSENSE, as well as work performed outside the project, has begun to tackle several crucial issues concerning SOTA AI vulnerabilities and limitations, aiming to develop a next-generation AI architecture with cognitive abilities – a system that is able to interact directly and in real-time with the real world.

IntuiCell logo