Artificial intelligence is often described in language that makes it sound like magic, models that understand, reason, and learn. Strip away the vocabulary however, and a neural network is nothing more than mathematics: multiply, add, compare, adjust, rinse & repeat.
This presentation peers under the AI hood. Using a problem small enough to hold in your head, we’ll train a simple neural network in a live demo. We start with a network that knows nothing and guesses randomly, then watch it learn. You will see how error is measured, how that error is pushed back through the network, and how heaps of tiny adjustments turn particles into patterns.
No calculus prerequisite, no framework foreknowledge, no black boxes. Just the machinery laid bare, doing something you can follow and verify with your own eyes. Come join us to comprehend the AI fundamentals of loss functions, gradient descent, and backpropagation.
Speaker Bio
Paul Speulstra, with 3 decades of enterprise IT experience, is an accomplished cybersecurity specialist, software developer, manager, and speaker. From the early days of the very first boot sector, file, and macro viruses to today’s threats combating APTs, BECs, AI phishing, and ransomware attacks, Paul has an extensive and diversified background in the security sphere. Prior front-line experience as a cybersecurity specialist and incident responder inside a Fortune 500 global cybersecurity operations center, Paul has firsthand knowledge in the research, prevention, detection, and remediation of cyber threats and malware. Paul, currently working for Bayside Solutions, Inc., a cybersecurity VAR based in Saint Petersburg Florida, is the inventor and programmer of Cyrebrium™, the cyber brain that powers BSI’s suspicious email analysis and inspection service known as Orqet™ (formerly PhishQueue™).