<-Foreword       Go to ToC       Scope ->

(Informative)

Technical Specification: Artificial Intelligence Framework (MPAI-AIF) V2.2 – in the following also called MPAI-AIF V2.2 or simply MPAI-AIF – provides a standard environment where AI Workflows (AIW) composed of AI Modules (AIM) are initialised, dynamically configured, executed and controlled. Some AIWs can be standardised by MPAI, i.e., they perform standardised functions, expose standard interfaces, and execute explicit computing workflows. Other AIWs can be proprietary, provided they expose the interfaces specified by MPAI-AIF. Developers can compete in providing AIF components – AIWs and AIMs – that have standard functions and interfaces. These may have improved performance compared to other implementations. AIMs can execute data processing or Artificial Intelligence algorithms and can be implemented in hardware, software, or in a hybrid hardware/software configuration.

The MPAI-AIF  specifies two Profiles: Basic and Security.

The Basic Profile has the following features:

  • Independence: An AIF Implementation does not depend on the Operating System.
  • Modularity: The architecture is component-based with specified interfaces.
  • Encapsulation: Component interfaces are abstracted from the development environment.
  • Access: An AIF Implementation can access validated Components in the MPAI Store.
  • Implementation: Component can be implemented as software only, hardware only, and hybrid hardware-software.
  • Execution: An AIF Implementation can be executed in local and distributed Zero-Trust architectures.
  • Interaction: An AIF Implementation can interact with other Implementations operating in proximity.
  • Machine Learning functionalities are supported.

The Security Profile inherits the functionalities of the Basic Profile. In addition, the APIs enable access to the following Trusted Services:

  • A selected range of cyphering algorithms.
  • A basic Attestation function.
  • Secure Storage as RAM, internal/external flash, or internal/external/remote disk.
  • Certificate-based Secure Communication.

with the following general conditions:

  • An AIF Implementation can execute only one AIW containing only one AIM that may be a Composite AIM whose components AIMs cannot access the Security API.
  • The AIF Trusted Services may rely on hardware and OS security features already existing in the hardware and software of the AIF environment.

Various actors – developers, integrators, and end users – benefit from the creation-com­pos­ition-execution-update of AIM-based work­flows interconnecting multi-vendor AIMs trained to specific tasks, operating in the standard AI framework and exchanging data in standard formats:

  • Technology providers can offer standard-conforming AI technologies to an open market
  • Application developers can find the technologies they need on the market.
  • Innovation is fuelled by demand for novel/ more performing AI components
  • Consumers have a wider choice of better AI applications from a competitive market
  • Society can lift the veil of opacity from large, monolithic AI-based applications.

An AIW and its AIMs may have three Interoperability levels:

  1. Level 1 – Proprietary and conforming to the MPAI-AIF Standard.
  2. Level 2 – Specified by an MPAI Application Standard.
  3. Level 3 – Specified by an MPAI Application Standard and certified by a Performance Assessor.

MPAI offers Users access to the promised benefits of AI with a guarantee of increased transparency, trust and reliability as the Interoperability Level of an Implementation moves from 1 to 3.

The chapters and sections of this Technical Specification are Normative unless they are labelled as Informative. Terms beginning with a capital letter are defined in Table 1 if specific to this MPAI-AIF Technical Specification. All MPAI-defined Terms are accessible online.

<-Foreword       Go to ToC       Scope ->