Breakfast and coffee served for all Train AI attendees on the 2nd Floor
Traditionally, software is built by programmers who consider the possible situations and write rules to deal with them. But recently, many applications have been created by machine learning: the programmer is replaced by a trainer, who shows the computer examples until it learns to complete the task. This shift in the way software is built is opening up exciting new possibilities and posing new challenges.
The biggest change to CrowdFlower’s business in the last year has come from deep learning and computer vision applications. The use cases range from health care to self-driving cars to drones to animal conservation. CrowdFlower has been hard at work building tools to support our customers labeling and human in the loop workflows.
Kaggle is platform for machine learning competitions, where companies post very difficult problems and data scientists compete to submit the best answer.. We have worked on problems ranging from predicting when somebody will have a seizure (based on EEG readings) to using algorithms to grade high school essays. This talk will draw on Kaggle’s experience to survey what’s possible at the limits of machine learning.
Adobe is changing the world of digital creativity through Adobe Sensei, a unified AI and machine learning framework that harnesses trillions of content and data assets, from high-resolution images to customer clicks. Sensei delivers deep understanding of creative content, from semantics to aesthetics, and increasingly drives the process of creating digital experiences. From inspiration, to discovery, to re-imagination and beyond, learn how Sensei is changing the world of digital creativity.
Five years ago, I gave a series of talks called “Where’s My Talking Robot”, which focused on the progress towards conversational AI, and the ways in which AI is already making more decisions about our lives than many people realize. That presentation made predictions about what the state of AI would be today, and so I will revisit AI in areas including the media, self-driving cars, education, health, and military defense. I will talk about how current technologies are supporting the development of conversational AI in a number of ways that directly impact people’s lives in many areas, and also about what the next five years might look like as technology grows.
In theory, AI algorithms can be applied to a wide set of problems of importance to humanity. In practice, however, technology itself is only a small portion of the overall success of an application. This talk discusses crucial issues that must be addressed in most successful applications of Artificial Intelligence. Key issues include quality, robustness, usability, optimal mix of humans and automation, trust, evolution, and economics.
The most common approaches to using machine learning often use humans to initially label a dataset for training and afterwards involve humans to check for quality. In this talk, Toby Segaran, engineering director at Reddit and author of “Programming Collective Intelligence” will explore ways that human intelligence can be used during the process of optimization, topic modeling and entity resolution to improve algorithmic results as they’re happening.
Presentations from six of CrowdFlower’s Customers on making AI work in the real-world.
Microsoft recently achieved a historical human parity milestone to recognize conversational speech on the switchboard task. Microsoft Cognitive Toolkit (CNTK) is the secret weapon that enabled this historical breakthrough. This talk will explain the story behind the scene how Microsoft did this with CNTK.
Bloomberg Beta’s Shivon Zilis and James Cham review their latest Machine Intelligence landscape—what successful MI companies are doing and what problems are open for startups.
Drawing from decades in data science, Monica will share war stories from the past, best practices for the present, and wild speculations about the future of data and machine learning.
A key ability of a high definition, “self-healing” map is being able to reflect changes in the real world, in real-time given the environment the vehicle is in. Leveraging machine learning technology, HERE can better provide vehicles on the road real-time change detection in the world, by comparing a vehicles environment model with HERE’s reference cloud based model to determine inconsistencies and thus determine changes in the real-world.
A popular goal of AI is to replace the need for human intervention and energy expense at any given task, but is this the most optimal model? The Kasparov approach, more commonly known as human-in-the-loop, increasingly seems like the best pathway to reach a high-achieving system. This talk discusses the implementation of these hybrid structures that take what works best in AI and humans, and create outperforming products.
AI has proven value in image analysis and other data intensive tasks relevant to medicine. With digital clinical imagery becoming ever more pervasive, AI is poised to improve the accuracy and efficiency of medical practice. A cooperative, rather than a competitive relationship with healthcare providers will be the key to adoption.
Deep learning has made great progress in a variety of language tasks. However, there are still many practical and theoretical problems and limitations. In this talk I will introduce solutions to some of these: How to predict previously unseen words at test time. How to have a single input and output encoding for words. How to grow a single model for many tasks. How to use a single end-to-end trainable architecture for question answering.