Al giorno d’oggi l’intelligenza artificiale necessita della spinta innovativa apportata dagli standard, come è avvenuto in passato per altre branche dell’information technology. Alcune organizzazioni non-profit si impegnano su questo fronte. Una di queste, MPAI (Moving Picture, audio and data coding by Artificial Intelligence) propone di superare gli attuali modelli di licenza e di creare un ecosistema in cui gli operatori possano progettare moduli interoperabili tra loro “by contract”. Contestualmente, l'IA deve diventare più etica, inclusiva e democratica, introducendo regole, parametri e requisiti trasparenti e universalmente condivisi. MPAI ha progettato MPAI-AIF con l'obiettivo di essere una spinta innovativa, capace di guidare la ricerca e l'industria dell'IA. L'analisi della specifica di MPAI-AIF e gli incontri di revisione con il “Development Committee” MPAI sono state le prime attività svolte in questa tesi; successivamente, è stata sviluppata una prima implementazione di MPAI-AIF per uno scenario IoT resource-constrained, in cui il dispositivo permette di verificare il corretto movimento riabilitativo di un arto rispetto a un pattern audio (ad esempio, la pulsazione di un metronomo). La prototipazione permette di verificare e validare la bontà dello standard e di individuare eventuali criticità. Il lavoro sul campo ha inoltre permesso di stilare una lista di “lessons learned”, verificando se l'approccio altamente modulare descritto da MPAI-AIF sia praticabile in contesti come l'IoT, dove la gestione di componenti hardware diversi tra loro rende difficile lo sviluppo di moduli standardizzati. Questi spunti faciliteranno anche futuri interventi e ottimizzazioni del prototipo, dal momento che esso farà parte di un progetto sponsorizzato dalla Eclipse Foundation e sarà il punto di partenza per successivi lavori nell'ambito dell’Aggregate Programming.
Artificial intelligence today needs the innovative push brought by standards, just as it has in the past for other areas of information technology. Some non-profit organisations join this cause. One of them, MPAI (Moving Picture, audio and data coding by Artificial Intelligence) proposes to overcome current licensing models and create an ecosystem in which implementers can design modules that are interoperable with each other by contract. At the same time, AI needs to be more ethical, inclusive and democratic by introducing transparent and universally shared rules, parameters and requirements. MPAI designed MPAI-AIF with the aim of being a driving force that can drive AI research and industry. Analysis of the specification and review meetings with the MPAI Development Committee were the initial activities carried out for this thesis; subsequently, I developed a first implementation of MPAI-AIF for a resource-constrained IoT scenario, in which the device allows to verify the correct rehabilitative movement of a limb with respect to an audio pattern (e.g., the pulse of a metronome). Prototyping makes it possible to verify and validate the soundness of the standard or to detect any bottlenecks. The fieldwork also allowed me to draw up a list of “lessons learned”, verifying whether the highly modular approach described by MPAI-AIF is practicable in contexts such as the IoT, where the control of different hardware components makes the development of standardised modules difficult. These suggestions will also facilitate future interventions and improvements of the prototype, as it will be part of a project sponsored by the Eclipse Foundation and will be the starting point for subsequent work in the area of Aggregate Programming.
Nuove architetture AI per l'IoT distribuita
BORTOLUZZI, DANIELE
2021/2022
Abstract
Artificial intelligence today needs the innovative push brought by standards, just as it has in the past for other areas of information technology. Some non-profit organisations join this cause. One of them, MPAI (Moving Picture, audio and data coding by Artificial Intelligence) proposes to overcome current licensing models and create an ecosystem in which implementers can design modules that are interoperable with each other by contract. At the same time, AI needs to be more ethical, inclusive and democratic by introducing transparent and universally shared rules, parameters and requirements. MPAI designed MPAI-AIF with the aim of being a driving force that can drive AI research and industry. Analysis of the specification and review meetings with the MPAI Development Committee were the initial activities carried out for this thesis; subsequently, I developed a first implementation of MPAI-AIF for a resource-constrained IoT scenario, in which the device allows to verify the correct rehabilitative movement of a limb with respect to an audio pattern (e.g., the pulse of a metronome). Prototyping makes it possible to verify and validate the soundness of the standard or to detect any bottlenecks. The fieldwork also allowed me to draw up a list of “lessons learned”, verifying whether the highly modular approach described by MPAI-AIF is practicable in contexts such as the IoT, where the control of different hardware components makes the development of standardised modules difficult. These suggestions will also facilitate future interventions and improvements of the prototype, as it will be part of a project sponsored by the Eclipse Foundation and will be the starting point for subsequent work in the area of Aggregate Programming.File | Dimensione | Formato | |
---|---|---|---|
767134_tesi_magistrale_bortoluzzi.pdf
non disponibili
Tipologia:
Altro materiale allegato
Dimensione
3.64 MB
Formato
Adobe PDF
|
3.64 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14240/85938