AI
Showing 40 out of 40 results
Software 2.0 and AI as an Engineering Discipline
We break down Stanford associate professor of computer science Chris Re’s recent talk on “Software 2.0” and “AI as an engineering discipline” and bring SVP of Unity Technologies Danny Lange into the conversation.
What Does It Take To Be a Data Scientist?
Data science is so much more than collecting, sorting and analyzing data. What does it take to be a data scientist and how does a day in the life of a data scientist look like? Ekaterina Sirazitdinova, Prayson Daniel and Nicholai Stålung will give you an insight into this and more.
On the Road to Artificial General Intelligence
Join this session to discuss the role of intelligence in biological evolution and learning. Danny Lange will demonstrate why a game engine is the perfect virtual biodome for AI’s evolution. Attendees will recognize how the scale and speed of simulations is changing the game of AI while learning about new developments in reinforcement learning.
Machine Learning on Source Code
source{d} is building the open-source components to enable large-scale code analysis and machine learning on source code. Their powerful tools can ingest all of the world’s public git repositories turning code into ASTs ready for machine learning and other analyses, all exposed through a flexible and friendly API. Francesc Campoy, VP of Developer Relations at [source{d}](https://sourced.tech/), will show you how to run machine learning on source code with a series of live demos.
Machine Learning with TensorFlow and Google Cloud
What is TensorFlow, why is it so popular, and how can you leverage it to build Machine Learning applications? We will walk through an end-to-end example including data ingestion, training, and prediction on a real dataset using a neural network. This talk will also cover how to use Google Cloud to supercharge your training and prediction as well as to remove pain from your development and operational workflows.
Software Engineering Principles First, Machine Learning Second
In the last decade, there is a huge hype to apply machine-learning techniques to support automation of almost any task during software development. Examples include application of machine learning techniques to debugging and automated program repair as well as automating major parts of the software testing process. However, analyzing the developed techniques shows that just throwing black-box machine learning techniques at software development problems will not provide the desired improvements that we expect. Consequently, a careful analysis of our software development practices is required, before we aim to automate them supported by tailor-made machine learning techniques. **prerequisite attendee experience level:** advanced
Machine learning in medical software done right
Machine learning and AI has reached the field of medical devices and medical software in an unprecedented manner, making workflows easier and more efficient for end users. Peter will talk about how to produce medically validated cloud-based software that is both effective and safe based on his experience in the industry.
Your Superpower User Manual
Did you know you have superpowers? The average person now has access to more data and more powerful tools and computing resources for analyzing that data than ever before in human history. In this tutorial, we'll show you how to combine Google cloud services (like Dataflow, BigQuery, and AutoML), along with modern Python tools to analyze and visualize some interesting and practical questions. You'll leave this session with a better sense of how much data analytic power is available right at your fingertips. **prerequisite attendee experience level:** beginner
Cloud operated Open Source AI Robots
There’s a lot going on in AI, and it’s a very fast moving space. This talk will cover how AI is being used to solve many interesting problems, and talk about the technologies available on AWS. There are easy to use high level services, high level Python frameworks like Keras and Gluon, and lower level frameworks like Tensorflow and Apache MXNet with more flexibility. Finally, we will take a look at the DIY Robocars project, an open source self driving race-car community, and the hack day competition that AWS is hosting at re:Invent 2017 in Las Vegas.
Emotion AI: A New Frontier
As human-machine interfaces become more prevalent in both home and work arenas, and our interactions with technology become more relational and conversational, it is imperative that we create machines that have a deep understanding of the user's intent. Emotions, through audio and visual cues, can enhance that understanding and lead to more relevant and personalized interactions. Emotion AI -- the intersection of human emotion analysis and artificial intelligence -- can provide the necessary information for machines to both learn and process these cues. The result transforms our interactions with apps and digital experience from clinical ones to empathetic ones. This is the promise of Emotion AI: machines that can learn, anticipate, and react to our emotions. It’s a technology that not only changes the way our devices see us, but how we see them. From social robots and autonomous cars, to education systems and healthcare chatbots, the convergence of human emotions and artificial intelligence is ushering a new level of computer interaction and paving the way for the Emotion AI economy. This presentation will explain: (1) how Emotion AI works, (2) how developers can integrate Emotion AI into the apps, devices, AI systems and digital experiences they are building, and (3) how Emotion AI is applied in real world solutions today. This will be demonstrated through live demos and many examples and uses cases covering a variety of verticals including healthcare, education, automotive, robotics, gaming and advertising.
The Meaning of (Artificial) Life
The Hitchhiker's Guide says the meaning of life is 42. Considering that the field of Data Science is going through a period of exponential growth it too could soon find that the meaning of an artificial life is also 42. But if you are not involved on a day-to-day basis, the expansion can seem bewildering. The story of how disparate disciplines have combined to produce Data Science is fascinating. In this talk, we will walk through a journey of scientific discovery. Following how, from humble beginnings, a multitude of sciences (and a surprising number of hacks) converged into the incredible advancements that you see in the media today. With these building blocks, we will be able to succinctly describe what these disciplines are and how they relate. The result will be the decomposition of a "rockstar" data science application; you will see that it is not so complicated after all. But the interesting result is that this generates a philosophical and political minefield; we can decompose the application and clearly see how it is built, but it also mimics or surpasses human capabilities. Are these human qualities? Is a more efficient or productive algorithm better than a human? Can we call them "intelligent"? Attendees will gain a fundamental understanding of the field of data science. You will leave understanding exactly the difference between machine learning and deep learning and how they are different. You will be able to describe how data mining can help your business run analytics tasks to improve efficiencies. You will be able to explain to your children why big data techniques were invented to solve a specific problem. This will suit anyone interested in the history of data science and also serve as a broad introduction to the rest of the day's in-depth talks. So, is the meaning of life 42? Possibly. But maybe all we need is a science algorithm to ask a better question.
Improving Business Decision Making with Bayesian Artificial Intelligence
In a world where deep learning and other massively scalable perception machines are at our disposal, allowing us to build amazing applications, the time is now ripe to move beyond the concept of pure perception and into broader Artificial Intelligence (AI). The path towards AI goes through what's missing in many applications today; Inference. Only when we combine Inference machines and Perception machines can we truly talk about AI. The benefit will be a machine that knows what to expect before observing it's environment and that can take prior information into account. With ever more mature Probabilistic programming languages available, we can express this marriage of perception and inference. In this talk we will scrape the surface of how to build Bayesian predictive inference machines using Probabilistic programming. Resources: Chapter 5 MCMC Handbook: https://arxiv.org/abs/1206.1901 Statistical Rethinking: https://www.amazon.com/Statistical-Rethinking-Bayesian-Examples-Chapman/dp/1482253445 Bayesian Data Analysis: https://www.amazon.com/Bayesian-Analysis-Chapman-Statistical-Science/dp/1439840954/ Stan, a probabilistic programming language: http://mc-stan.org/
One does not simply put Machine Learning into Production
When deciding to infuse existing products with machine-learning smarts, or building ML-first products, there are multiple challenges to be aware of. First, you and your organization need to understand important dimensions -- accuracy, cost, maintainability, interpretability -- and trade-offs between them. Second, several technical challenges present themselves when deploying data science experiments into production environments. I will share some lessons learned while building ML products serving billions of predictions to live customers -- and hopefully provide some take-aways for anyone in the audience looking to indeed put machine learning into production.
Improving Business Decision Making with Bayesian Artificial Intelligence
In a world where deep learning and other massively scalable perception machines are at our disposal, allowing us to build amazing applications, the time is now ripe to move beyond the concept of pure perception and into broader Artificial Intelligence (AI). The path towards AI goes through what's missing in many applications today; Inference. Only when we combine Inference machines and Perception machines can we truly talk about AI. The benefit will be a machine that knows what to expect before observing it's environment and that can take prior information into account. With ever more mature Probabilistic programming languages available, we can express this marriage of perception and inference. In this talk we will scrape the surface of how to build Bayesian predictive inference machines using Probabilistic programming. **Resources**<br /> [Chapter 5 MCMC Handbook](https://arxiv.org/abs/1206.1901)<br /> [Statistical Rethinking](https://www.amazon.com/Statistical-Rethinking-Bayesian-Examples-Chapman/dp/1482253445)<br /> [Bayesian Data Analysis](https://www.amazon.com/Bayesian-Analysis-Chapman-Statistical-Science/dp/1439840954/)<br /> [Stan, a probabilistic programming language]( http://mc-stan.org/)
Human-Computer Partnerships
Incredible advances in hardware have not been matched by equivalent advances in software; we remain mired in the graphical user interface of the 1970s. I argue that we need a paradigm shift in how we design, implement and use interactive systems. Classical artificial intelligence treats the human user as a cog in the computer's process -- the so-called “human-in-the-loop”; Classical human-computer interaction focuses on creating and controlling the 'user experience'. We seek a third approach -- a true human-computer partnership, which takes advantage of machine learning, but leaves the user in control. I describe a series of projects that illustrate our approach to making interactive systems discoverable, appropriable and expressive, using the principles of instrumental interaction and reciprocal co-adaptation. The goal is to create robust interactive systems that significantly augment human capabilities and are actually worth learning over time.
Beyond AI
<h3>Using games to create human-machine intelligence</h3> Artificial intelligence (AI) is a field in rapid development and there is almost daily news reports about how AI has revolutionized yet another industrial domain. Some researchers claim that within a few decades we will reach a so-called singularity in which computer intelligence will surpass human capabilities in all domains. Other AI-researchers, however, maintain that we have still far from understood the human ability to reach fast, intuitive and correct decisions based on often seemingly too little data. They believe that the future job market will rely heavily on such insights in developing hybrid forms of human-machine intelligence. In the [scienceathome.org](https://www.scienceathome.org/) project, we have developed games allowing so far 250,000 players to contribute to research in quantum and classical physics, mathematics, chemistry, behavioral economics, corporate management, psychology and cognitive science. We believe that this wealth of data from human individual and collective problem solving holds the key to understanding and further exploiting human uniqueness. Apart from AI, the digital 21st century also poses a democratic challenge through the increasing concentration of large data sets of human behavior among a few high-tech companies. As a response, we have recently launched a game-based effort aimed at cognitively profiling a large portion of the world’s population and making the dataset available for open scientific research. In a future, where knowledge is power, we believe that the best possible society can only created if we achieve equality in the access to knowledge of human behavior.
Artificial Intelligence Reloaded - AI Applications in the Industry
Bringing Artificial Intelligence applications to life is much more than running an AI framework on an artificial data set. It starts with data gathering, consolidation, cleaning and continues with data science and model building. It goes all the way until deployment, DevOps and lifecycle management. In this talk you'll learn on how AI will change the today software development and deployment processes. **Prerequisite attendee experience level:** professional
Why Algorithmic Fairness Needs Democracy
Our lives are increasingly structured by algorithms. However, a number of well-publicized incidents have raised awareness of the discriminatory power of algorithms (e.g. Google Photos mistakenly labelled black people as “gorillas”, the COMPAS recidivism algorithm is more likely to misclassify African Americans as reoffenders than white Americans, etc.). To the credit of the engineering and research community, efforts have been made to combat discrimination via algorithms and to achieve algorithmic fairness. Yet, I show why algorithm developers and researchers alone could not achieve algorithmic fairness, and point out that it can only be achieved together with the public. In other words, algorithm developers and researchers must work with the public if they want algorithms to be fair, or simply—algorithmic fairness needs democracy. **prerequisite attendee experience level:** Beginner
Wetware Intelligence
As ambitions to create newer and faster supercomputers grow, so do the challenges. Increasing computational power comes with demands of scale, stability, and accessibility. In his keynote, Osh, CEO and Founder of Koniku, will tell us how they are working to solve this by harnessing the power of biological neurons to create the next generation supercomputers . "Real" neurons are connected, not just to each other but to silicon chips. At first those chips are being used in devices that sense airborne chemicals. Other possible applications await the technology in industries from drug development to agriculture. A successful brain-machine connection will help us study and treat neurological diseases. The greatest ambition of them all is to create wetware artificial intelligence. Wetware is no longer science-fiction. Combining biology and machine lets us perform tasks that neither one of the two could ever do alone.
The Promise and Limitations of AI
Almost everyone who talks about Artificial Intelligence, nowadays, means training multi-level neural nets on big data. Developing and using those patterns is a lot like what our right brain hemispheres do; it enables AI's to react quickly and – very often – adequately. But we human beings also make good use of our left brain hemisphere, which reasons more slowly, logically, and causally. I will discuss this "other type of AI" – i.e., left brain AI, which comprises a formal representation language, a "seed" knowledge base with hand-engineered default rules of common sense and good domain-specific expert judgement written in that language, and an inference engine capable of producing hundreds-deep chains of deduction, induction, and abduction on that large knowledge base. I will describe the largest such platform, Cyc, and will demo a few commercial applications that were produced just by educating it as one might teach a new human employee. But it is important to remember that human beings' "super-power" is our ability to harness both types of reasoning, and I believe that the most powerful AI solutions in the coming decade will likewise be hybrids of right-brain-like "thinking fast" and left-brain-like "thinking slow". That is the only path I see by which we will overcome the current dangerous inability of deep-learning AI's to rationalize and explain their decisions, and will make AI's far more trusted and – more importantly – far more trustworthy. Anyone who understood this abstract and found it interesting should find my actual talk similarly accessible – and hopefully interesting!
Computers are Stupid: Protecting "AI" from Itself
It seems like everywhere you turn there is another startup or company looking to use "AI" to revolutionize something, or really anything. Is AI hype or substance? In this talk, we'll dive into AI security, looking at the field of adversarial learning. How easy is it to fool an artificial intelligence? What would be needed to create a robust and secure neural network? How are researchers working on solving the security issues within the way we train AI, to help it from making errors or being used for unethical tasks. In that same vein, we'll address how machine learning places data privacy and ethical data use at risk. We'll explore why new efforts like GDPR and privacy-preserving ML might make a way for a safer, more ethical machine learning practice.
Build a Q&A Bot with DeepLearning4J
Chatbots are here - you no longer necessarily talk to a human when you contact your insurance agency. Whether that's a good thing remains to be seen, but it sure is interesting for us as developers. The primary goal of my talk is to show you how you can use DeepLearning4J to build a neural network for answering frequently asked questions. I will show you how to build, train, test and use a the neural network in a basic chatbot. This talk is aimed at developers who have heard of neural networks, but don't want to get involved in all the math behind it. This is a not-so-scientific introduction into the wonderful world of chatbots and AI. **Prerequisite attendee experience level:** beginner
Inextricably Linked: Reproducibility and Productivity in Data Science and AI
Because it is more complex and has far more moving parts, Data Science & AI is where Software Development was in 1999: people are emailing and Slacking notebooks to each other, due to a lack of appropriate tooling. There are few CI/CD pipelines and model health monitoring is scarce. A lot that could be automated is still manual. And teams are siloed. This causes problems both for productivity: it's hard to collaborate, and reproducibility: which impacts on governance and compliance. In this talk, Mark shares his team’s research comparing the evolution of Software Development & DevOps with that of Data Science & AI. Mark then presents a proposal for an architecture and a set of open source tools to solve both the collaboration and the governance problem in Data Science & AI. With live demos!
Code + AI: Will Robots Take Our Coding Jobs?" Machine Learning Applied to Programming
Machine learning is permeating every facet of our lives, from learning our preferences to self-driving cars, but what happens when you apply neural networks to code? How do you even view code as data? The key ideas are easy to summarize and fun to play with. This talk will provide an overview of fundamental concepts of machine learning, and then delve into how learning can be used to analyze and improve code. The talk will also provide pointers to available commercial and open source tools and discuss what’s been achieved so far (coding in English, context-aware code completion, automated Stack Overflow). The talk will close with speculation on where the field is going, and how machine learning won’t take our jobs, but hopefully will take over some of the repetitive work we don’t like doing.
The Mechanized Pen
Ross is not a poet. But he will discuss his experiences with writing machines and share lessons learned along the way working on projects including: * expressive cameras * computer-generated screenplays * a day glo poetic lion statue * and a novel written with a car These and other experiments and provocations are all centered around developing new forms and interfaces for written language, enabled by machine intelligence. **Who should attend this talk:** Anyone interested in computational creative writing. **Academic level:** All levels / not applicable, as my talk is more about applying code to creative pursuits than academic CS. **What is the take away in this talk:** Consideration of new interfaces and forms for creativity, enabled by machine intelligence.
TensorFlow Lite: how to Accelerate your Android and iOS App with AI
TensorFlow Lite is TensorFlow’s lightweight solution for Android, iOS and embedded devices. It enables on-device machine learning inference with low latency and a small binary size. TensorFlow Lite also supports hardware acceleration with the Android Neural Networks API and Apple Core ML. In this session, we will discuss how developers can use TensorFlow Lite to overcome the challenges for bringing the latest AI technology to production mobile apps and embedded systems.
The Fast Track to AI with Serverless
Join us for this interesting session *"The Fast Track to AI with Serverless"* with Peter Elger. Until recently, adopting and applying AI and machine learning capabilities in a software platform or a typical enterprise technology estate was out of the reach of most developers and required highly skilled experts. Lately, we have seen rapid growth in the range and capability of cloud-native AI services from all the major providers. **Armed with a basic understanding of the underlying concepts, developers can now adopt machine learning tools to solve real-world business problems and add advanced features to their platforms without needing a multi-year research project.** This talk will be based on Peter's book [AI as a Service](https://www.manning.com/~i-as-a-service), published by Manning. Focusing on Node.js and the AWS stack, this talk will cover the range and scope of services available off the shelf today and these services can be adopted by developers through familiar API interfaces. Moreover, Peter will discuss patterns for the adoption of AI services that can be used to augment existing systems and platforms with AI capabilities. The talk will include a discussion of several example systems with code and live demonstrations. It will also review some real-world projects and share experiences of using AI services in the wild. **In this talk, you'll learn:** * Key points from Peter's book * Which services that are available today regarding Node.js and AWS stack and much more about these services
Best Practices for Real-time Intelligent Video Analytics
**Katja is a research engineer focusing on deep learning, computer vision and inference optimization**. Larger and more complex Vision AI networks enable better accuracy and precision since they are able to encode more information. This increase in size and complexity is, in turn, naturally associated with trained AI models having lower throughput and larger memory requirements. With that, real-time AI inference is becoming a new great challenge in intelligent video analytics. NVIDIA’s approach at solving this problem relies on two major components: first, tuning AI models for performance depending on the target deployment hardware platform, and, second, optimizing the use of available GPUs. From this talk, **you will learn how to leverage this approach to achieve real-time inference performance** by using software tools like DeepStream SDK, TensorRT and Triton Inference Server.
Who's Afraid of the Black Box Models?
**Prayson is on the mission to help companies gain AI competitive advantage by growing revenue, slashing production timelines, multiplying efficiency and making data-driven decisions.** Take a dive into the deep depths and pitfalls of explainable machine learning, going beyond the illusions of interpretability and explainability. Draw from Prayson's experience and explore how ethical data handling and counter-factual fairness model testing help in keeping black-box models and yet satisfy GDPR and ALTAI ([Guidelines for Trustworthy AI](https://digital-strategy.ec.europa.eu/en/library/communication-building-trust-human-centric-artificial-intelligence))
Abzu’s Pioneering Technology, the QLattice®, Introduces a New Standard of Interpretability to Artificial Intelligence
**Jaan will introduce a new standard of interpretability to artificial intelligence.** He is passionate about the way AI, advanced data analysis and adaptive/responsive systems are poised to revolutionize the world at a scale not seen since the industrial revolution. In this presentation, Jaan will explain the core technology of the QLattice, which was inspired by physicist Richard Feynman’s work on path integrals. He’ll demonstrate the power of searching through, in theory, an infinite number of possible models to find the best fit and reveal the simple mathematical equation that is the solution to your problem. Jaan will be using Abzu’s Python library Feyn® in a Jupyter notebook to explore a synthesized dataset with real-world application.
How Corsearch Manages Software Delivery at Scale
Software is integral to the pace of innovation at almost every business in the world. Yet as companies scale the culture and technology used to build, test and deploy software often struggles to keep up. In this session, Harness and Corsearch will share the latest developments and trends in software development. With a focus on how new technologies, powered by machine learning and AI-driven, are improving developer experiences and increasing productivity.
Large Language Models: Friend, Foe, or Otherwise
Alex Castrounis is the founder and CEO of Why of AI, a best-selling book author on AI, and a professor of AI for Northwestern University's Kellogg / McCormick MBAi program. He has over two decades of experience advising startups to Fortune 100 companies on using data, analytics, and AI models to drive business growth and customer success. Alex is also a former INDYCAR engineer, race strategist, and data scientist that drove winning results for teams such as Andretti Autosport in world-renowned races, including the Indy 500. His experience applying analytics to make smarter, data-driven decisions in racing now helps businesses gain and maintain a competitive advantage in their industry.
Abzu’s AI Engineering Journey: How the QLattice® Went From Freestanding Algorithm to SaaS Discovery Engine
The concept behind a new class of AI had been brewing in Casper’s head for nearly 30 years when he and six curious nerds founded the Danish/Spanish startup Abzu in 2018. The QLattice launched in April 2020 and introduced a new standard of interpretability to artificial intelligence. Three years later, the QLattice has revealed numerous insights in life science and drug development and is used in the largest pharmaceutical and most ambitious biotech companies in the world. Today, Abzu is building Reason®, a user-friendly discovery platform built on the QLattice discovery engine, to make the QLattice’s accelerated exploration accessible to many more scientists and researchers. In this presentation, Casper will recount Abzu’s successes – and misses – in its AI engineering journey and demonstrate real-world applications of the pioneering QLattice.
A Composer’s Guide to Creating with Generative Neural Networks
While generative text and image models like ChatGPT and DALL-E2 have garnered popular attention, generative music creation has only begun to blossom. This talk will focus on a case study of an independent composer navigating the infrastructure and coding challenges of building, optimizing, and training a deep learning model for audio generation for use in her own compositions. Molly Jones, a composer and software engineer, attempts to develop creative work that is aesthetically, conceptually, and ethically sound while addressing problems of data augmentation, solo coding, and limited resources. The case study does not offer immediate solutions to the problems faced by independent artists but rather a catalog of areas for future research aimed at easing the use of deep learning models in creative contexts. Molly is an Associate Software Consultant at Spantree, LLC, and a Composer in Residence at the University of Toronto, where she explores the intersections of AI and music.
Lunch & Roundtable Discussions
Don't miss your chance to share, learn, and network with speakers and attendees! Grab your lunch and join us at one of the roundtables. **Teams & Inclusion** Join Bryan Cantrill and Sara Caldwell in a vibrant conversation about building inclusive, high-performing teams in today's diverse tech landscape. Share your experiences and learn from others on fostering collaboration and embracing diversity in the workplace. **Programming with AI** Unlock the potential of AI in software development with Alex Castrounis and Linda Rising. Delve into the intersection of programming and artificial intelligence, discussing techniques, tools, and breakthroughs. Engage in thought-provoking conversation and exchange ideas. **Data Streams** Embark on a deep dive into data streams with Mary Grygleski and Kasun Indrasiri. Share and explore the latest in streaming technologies, architectures, and use cases. Share your experiences and learn from others in this dynamic roundtable on the cutting edge of data processing.
Web3 Beyond Blockchains
If you thought Web3 was all about cryptocurrencies and public ledgers, think again. A new generation of infrastructure is emerging that uses consensus, verification and transparency to solve some of the biggest shortcomings in enterprise computing today. Join Cornel Montano, CTO at Forte Group, Jason Teutsch, Founder of Truebit, and a panel of experts who will discuss how the futures of serverless compute, the composable web and large-scale AI are all converging, with Web3 as the common denominator. One critical aspect we will explore is the criticality of engaging with the right development partner when kick-starting bleeding technology initiatives.
“Hello Deep Learning”
Deep learning/generative AI is taking the world by storm. Until late 2022 I had mostly dismissed all of this as hype, but the advent of ChatGPT made it impossible to ignore this any longer. By now there are tons YouTube videos telling you how you too can script up "an AI". If you feel like I did in 2022 that you must catch up rapidly, these videos are not the way to go. I have created a "Hello Deep Learning" project which explains the essence of deep learning absolutely from scratch. By this I mean an empty directory. In this presentation we'll build an OCR solution that can read actual handwritten letters, absolutely from scratch. The goal of "hello deep learning" is to get in on the ground floor so you can start learning more, while actually knowing what is going on. This talk introduces “Hello Deep Learning”, and also covers the fascinating actual basics of machine learning - stuff that most deep learning users have no idea about. This knowledge should help you find your way in the brave new world of AI.
Alex Castrounis
Professor of AI for Northwestern University's Kellogg / McCormick MBAi program, CEO of Why of AI
Why of AI
