January 31, 2020.
Top 5 Agenda Items at Shift AI 2020
With great speakers come great talks. AI industry is on the brink of its biggest revolution yet and there’s a lot of burning topics that need to be discussed. Theories of where the AI should move next and how it should move along the way will be discussed by the leading experts in the industry (some of which we’ve mentioned in our other article
). These are just some of the topics that will be covered (check out the full agenda here
5. Mission Collision: Artificial intelligence in the era of social justice reforms
Kairan Quazi - AI Summer Research at Intel
A daily scan of news headlines shows that technology has become invasive and pervasive in every aspect of modern civil life. The increasing use of artificial intelligence technology by law enforcement and criminal justice institutions is raising serious issues for civil liberties and social justice.
"Bad data" disproportionately targets already marginalized populations that have faced historical systemic discrimination. This discussion will touch on the consequences of the current application of Artificial Intelligence technology on initiatives to achieve equity and justice in our civil institutions.
4. Zero-Trust in a world of Deep Fakes: Challenges to AI-based Threat Detection
Sohrob Kazerounian - AI Research Lead at Vectra
In recent years, AI and automation have rapidly expanded the scope of what computers are capable of achieving. Speech recognition, computer vision, and natural language processing (though not understanding) are beginning to approach human-level performance. Nevertheless, there are a number of unique challenges that make the application of these technologies to cybersecurity a non-trivial task.
Most importantly, the adversarial nature of attackers means that threat detection must constantly be on the lookout for never before seen attacks that occur against a constantly evolving background of network infrastructure and traffic. By drawing analogies between the zero-trust framework in information security, and generative adversarial networks in AI, this talk will explore these challenges and approaches to overcoming them.
3. Ethically high risk AI applications -- proceed with caution (if at all)
Kathy Baxter - Ethical AI Practice at Salesforce
AI has the potential for significant good in the world; however, there are specific AI applications that are at high risk of directly or indirectly violating human rights. It is important to recognize why these applications are high risk, determine if the risks can be mitigated, and when or if they should be used at all.
2. Operationalizing Machine Learning in Enterprises: How companies can manage the entire data science process and productively use machine learning with ease
Sebastian Wieczorek - Vice President of Machine Learning at SAP
Machine Learning made large progress in recent time, and it has an enormous potential to revolutionize and simplify the daily work in companies. However, according to Gardener, this potential is still not yet fully exploited, and the biggest issue for companies nowadays is to deploy machine learning in production and make really use of it for their productive processes.
In this session Sebastian will show how SAP manages development, training, deployment, and operation of Enterprise AI applications for building the intelligent Enterprise.
1. Evolution of Data Science & ML/AI in Tech: from Insights to Experiment-Backed Data Products
Lin Huang - Director of Data Science & Engineering at Reddit
Success of data science requires mastery of ML/AI techniques, and deep understanding of company priorities. The most rewarding yet challenging aspect is to build a rhythm to deliver actionable outputs (i.e., data products) that directly transform business decisions. In tech companies, Lin and his team are blessed with uniquely complex data assets, for which it took Reddit rounds of iterations to customize a solution that maximizes their data potential.
This talk focuses on Lin’s journey as a leader and practitioner in the tech industry in the last decade, especially the last 18 months at Reddit to build an experimentation platform, and to revamp the engineering workflow prioritizing the interplay of offline ML modeling and online testing.
Not every piece of ML insights checks out in an experiment, but every "setback" is equally valuable as it allows them to iterate faster and more purposefully. Culturally speaking, Lin encourages calculated risk taking since data products are most powerful when they point to new business directions (vs. validating the existing ones).
With a suite of concrete examples, he demonstrates how rhythmically outputting data products answers the common questions that almost all serious tech companies are curious about, namely, where are “good ideas” from? How to generate them at scale? Do they hold up IRL? Still valuable if they don’t?