BEGIN:VCALENDAR VERSION:2.0 PRODID:IEEE vTools.Events//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Europe/Rome BEGIN:DAYLIGHT DTSTART:20260329T030000 TZOFFSETFROM:+0100 TZOFFSETTO:+0200 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3 TZNAME:CEST END:DAYLIGHT BEGIN:STANDARD DTSTART:20261025T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10 TZNAME:CET END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20260724T073402Z UID:5F8E617A-EF1C-45EF-A494-DBC309A6C4EB DTSTART;TZID=Europe/Rome:20260723T110000 DTEND;TZID=Europe/Rome:20260723T120000 DESCRIPTION:IEEE ComSoc Italy Chapter Distinguished Lecture Tour\n\nSpeaker : Prof. Carlo Fischione (KTH\, IEEE ComSoc Distinguished Lecturer)\n\nTitl e of Talk: AI in Wireless Networks: What Works\, What Doesn't\, and What C omes Next\n\nAbstract: Artificial Intelligence and Machine Learning are wi dely seen as key enablers of future wireless systems\, with promises rangi ng from fully autonomous networks to dramatic gains in spectral efficiency and energy consumption. However\, many of these expectations are based on implicit assumptions\, such as abundant data\, reliable connectivity\, an d centralized computation\, that do not hold in practical radio environmen ts. In this talk\, we critically examine the myths and realities of AI/ML in wireless networks\, with a focus on what can be realistically achieved within the constraints of modern radio systems. We discuss why the straigh tforward application of data-driven methods often falls short in distribut ed\, bandwidth-limited\, and latency-constrained settings\, and highlight common pitfalls in current approaches to learning over wireless infrastruc tures. Building on this analysis\, we motivate the co-design of communicat ion and computation\, where the wireless channel is not just a medium for data transfer\, but an active component of distributed learning and infere nce. Leveraging recent advances in over-the-air computation and its digita l implementations (e.g.\, ChannelComp)\, we outline how AI-native radio sy stems can emerge by embedding learning objectives directly into the physic al layer.\n\nCo-sponsored by: University of L'Aquila\n\nSpeaker(s): Carlo Fischione\n\nDipartimento di Ingegneria e Scienze dell'Informazione e Mate matica\, L'Aquila\, Abruzzi\, Italy LOCATION:Dipartimento di Ingegneria e Scienze dell'Informazione e Matematic a\, L'Aquila\, Abruzzi\, Italy ORGANIZER:daniele.tarchi@unifi.it SEQUENCE:12 SUMMARY:AI in Wireless Networks: What Works\, What Doesn't\, and What Comes Next URL;VALUE=URI:https://events.vtools.ieee.org/m/568365 X-ALT-DESC:Description: <br /><p><strong>IEEE ComSoc Italy Chapter Distingu ished Lecture Tour</strong></p>\n<p><strong>Speaker:</strong> Prof. Carlo Fischione (KTH\, IEEE ComSoc Distinguished Lecturer)</p>\n<p><strong>Title of Talk:</strong> AI in Wireless Networks: What Works\, What Doesn't\, an d What Comes Next</p>\n<p><strong>Abstract:</strong> Artificial Intelligen ce and Machine Learning are widely seen as key enablers of future wireless systems\, with promises ranging from fully autonomous networks to dramati c gains in spectral efficiency and energy consumption. However\, many of t hese expectations are based on implicit assumptions\, such as abundant dat a\, reliable connectivity\, and centralized computation\, that do not hold in practical radio environments. In this talk\, we critically examine the myths and realities of AI/ML in wireless networks\, with a focus on what can be realistically achieved within the constraints of modern radio syste ms. We discuss why the straightforward application of data-driven methods often falls short in distributed\, bandwidth-limited\, and latency-constra ined settings\, and highlight common pitfalls in current approaches to lea rning over wireless infrastructures. Building on this analysis\, we motiva te the co-design of communication and computation\, where the wireless cha nnel is not just a medium for data transfer\, but an active component of d istributed learning and inference. Leveraging recent advances in over-the- air computation and its digital implementations (e.g.\, ChannelComp)\, we outline how AI-native radio systems can emerge by embedding learning objec tives directly into the physical layer.</p> END:VEVENT END:VCALENDAR