<div dir="ltr"><div><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif"><br class="gmail-Apple-interchange-newline">[* Apologies in case of multiple posting *]</font></span></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">The Italian Association for Artificial Intelligence, in collaboration with the AISB (Society for the Study of Artificial Intelligence and Simulation of Behaviour, UK) is pleased to announce the next seminar of its Spotlight Seminars on AI initiative:</font></span></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">March 21  – 5:00 P.M. (CET)</font></span></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">Title: </span><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">Towards Verification of Neural Systems</span></font></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Speaker: Alessio Lomuscio, Imperial College London, UK</font></span></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">Link Streaming: </span><a href="https://www.youtube.com/@AixiaIt/streams" target="_blank"><span style="background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">https://www.youtube.com/@AixiaIt/streams</span></a><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"> </span></font></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Abstract:  </font></span></p><p dir="ltr" style="line-height:1.2;text-align:justify;margin-top:0pt;margin-bottom:0pt"><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">A major challenge in deploying ML-based systems, such as ML-based computer vision, is the inherent difficulty in ensuring their performance in the operational design domain. The standard approach consists in extensively testing models against a wide collection of inputs. However, testing is inherently limited in coverage, and it is expensive in several domains. Novel verification methods provide guarantees that a neural model meets its specifications in dense neighbourhood of selected inputs. For example, by using verification methods we can establish whether a model is robust with respect to infinitely many re-illumination changes, or particular noise patterns in the vicinity to an input. Verification methods can also be tailored to specifications in the latent space and establish the robustness of models against semantic perturbations not definable in the input space (3D pose changes, background changes, etc). Additionally, verification methods can be paired with learning to obtain robust learning methods capable of generating models inherently more robust than those that may be derived with standard methods. In this presentation I will succinctly cover the key theoretical results leading to some of the present ML verification technology, illustrate the resulting toolsets and capabilities, and describe some of the use cases developed with our colleagues at Boeing Research, including centerline distance estimation, object detection, and runway detection. I will argue that verification and robust learning can be used to obtain models that are inherently more robust than present learning and testing approaches, thereby unlocking the deployment of applications in society critical applications.</font></span></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><br></font></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><br></font></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><br></font></p><p dir="ltr" style="line-height:1.2;text-align:justify;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">Bio: </span><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">Alessio Lomuscio (</span><a href="http://www.doc.ic.ac.uk/~alessio" target="_blank"><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">http://www.doc.ic.ac.uk/~alessio</span></a><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">) is Professor of Safe Artificial Intelligence at Imperial College London (UK), where he leads the Safe AI Lab (</span><a href="http://sail.doc.ic.ac.uk/" target="_blank"><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">http://sail.doc.ic.ac.uk/</span></a><span style="font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline">). He is a Distinguished ACM member, a Fellow of the European Association of Artificial Intelligence and currently holds a Royal Academy of Engineering Chair in Emerging Technologies. He is founding co-director of the UKRI Doctoral Training Centre in Safe and Trusted Artificial Intelligence. Alessio's research interests concern the development of verification methods for artificial intelligence.  Since 2000 he has pioneered the development of formal methods for the verification of autonomous systems and multi-agent systems, both symbolic and ML-based. He has published over 200 papers in leading AI and formal methods conferences and journals. He is the founder and CEO of Safe Intelligence, a VC-backed Imperial College London spinout helping users build and assure robust ML systems.</span></font></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><font face="arial, sans-serif"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"></span></font></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">The joint AI*IA/AISB Committee of the Spotlight Seminars on AI, </font></span></p><font face="arial, sans-serif"><br></font><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Giuseppe De Giacomo</font></span></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Floriana Grasso</font></span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Antonio Lieto</font></span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Bertie Muller</font></span></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif">Luciano Serafini</font></span></p><p dir="ltr" style="line-height:1.2;margin-top:0pt;margin-bottom:0pt"><span style="color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;font-variant-alternates:normal;vertical-align:baseline"><font face="arial, sans-serif"><br></font></span></p></div><div><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div style="color:rgb(34,34,34)"><font color="#000000"><span style="font-size:12.8px">-------------</span><br style="font-size:12.8px"><font face="arial, helvetica, sans-serif">Antonio Lieto, Ph.D</font></font></div><div style="color:rgb(34,34,34)"><font color="#000000"><font face="arial, helvetica, sans-serif">Associate Professor of Computer Science at the University of Salerno, DISPC, Italy</font></font></div><span style="color:rgb(0,0,0);font-family:arial,helvetica,sans-serif">Cognition Interaction and Intelligent Technologies Lab (CIIT Lab) @ Unisa</span><div><span style="color:rgb(0,0,0);font-family:arial,helvetica,sans-serif"><a href="https://www.ciitlab.org" target="_blank">https://www.ciitlab.org</a>  </span><span style="font-family:arial,helvetica,sans-serif;color:rgb(0,0,0)"> </span><div style="color:rgb(34,34,34)"><font color="#000000"><font face="arial, helvetica, sans-serif">Research Associate at ICAR-CNR, Palermo, Italy<br></font></font></div><div style="color:rgb(34,34,34)"><font face="arial, helvetica, sans-serif" color="#000000">Home: <a href="https://www.antoniolieto.net/" style="color:rgb(17,85,204)" target="_blank">https://www.antoniolieto.net</a></font></div><div style="color:rgb(34,34,34)"><font color="#000000">ACM Distinguished Speaker: <a href="https://speakers.acm.org/speakers/lieto_12489" style="color:rgb(17,85,204)" target="_blank">https://speakers.acm.org/speakers/lieto_12489</a> <br></font></div><div style="color:rgb(34,34,34)"><div><br></div><div><div><font face="arial, helvetica, sans-serif" color="#000000">Deputy Editor in Chief of JETAI (Journal of Exp. & Theor. Artif. Intell., T&F)</font></div><div><font face="arial, helvetica, sans-serif" color="#000000"><a href="https://www.tandfonline.com/action/journalInformation?show=editorialBoard&journalCode=teta20" style="color:rgb(17,85,204)" target="_blank">https://www.tandfonline.com/jetai</a> </font></div></div></div><div style="color:rgb(34,34,34)"><div><font face="arial, helvetica, sans-serif" color="#000000">Associate Editor of Cognitive Systems Research (Elsevier)</font></div><div><font color="#000000"><a href="https://www.journals.elsevier.com/cognitive-systems-research/editorial-board" style="color:rgb(17,85,204)" target="_blank">https://www.journals.elsevier.com/cognitive-systems-research/editorial-board</a> </font></div></div><div style="color:rgb(34,34,34)"><font color="#000000"><br></font></div><div style="color:rgb(34,34,34)"><div><font color="#000000">Author of "Cognitive Design for Artificial Minds" (2021), Routledge, Taylor & Francis.</font></div><div><a href="https://www.amazon.com/dp/1138207950" style="color:rgb(17,85,204)" target="_blank">https://www.amazon.com/dp/1138207950</a></div></div></div></div></div></div></div>