=Paper=
{{Paper
|id=Vol-2269/keynote1
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-2269/keynote1.pdf
|volume=Vol-2269
}}
==None==
<pdf width="1500px">https://ceur-ws.org/Vol-2269/keynote1.pdf</pdf>
<pre>
                           Keynote: AI Canonical Architecture and Robust AI


                                                           David R. Martinez∗




                            Abstract
  This presentation addresses an AI canonical architecture suit-
  able for a number of different classes of applications. Several
  examples will be shown focused on cyber security and poten-
  tial vulnerabilities to adversarial attacks. One critical element
  of the end-to-end AI architecture is the need for robust AI.
  Significant advances have been made in AI algorithms and
  high performance computing. However, additional advance-
  ments in science and technology (S & T) are needed to vali-
  date the performance of AI systems. This performance assess-
  ment is very critical because AI systems are very brittle to ad-
  versarial modifications to the system. The AI canonical archi-
  tecture starts with data conditioning, followed by classes of
  machine learning algorithms, human-machine teaming, mod-
  ern computing, and robust AI. We will briefly address each of
  these areas. The presentation concludes with a summary of S
  & T challenges and recommendations.




    ∗
      D. Martinez is with the Lincoln Laboratories, Massachusetts
Institute of Technology, MA, USA. e-mail: dmartinez@ll.mit.edu
Copyright ⃝  c by the paper’s authors. Copying permitted for private
and academic purposes. In: Joseph Collins, Prithviraj Dasgupta,
Ranjeev Mittu (eds.): Proceedings of the AAAI Fall 2018 Sympo-
sium on Adversary-Aware Learning Techniques and Trends in Cy-
bersecurity, Arlington, VA, USA, 18-19 October, 2018, published
at http://ceur-ws.org

</pre>