logo

2022-10-19 ~ 2022-10-21

Presentations (with video): 59 (35)

Join Scale AI as we bring together over 120 of the world's brightest AI leaders, visionaries, practitioners, and researchers across industries to explore operationalizing AI and Machine Learning. This year's conference will bring together 30,000 AI leaders and practitioners and feature three days of in-person and virtual keynote presentations, fireside chats, expert panel discussions, and hands-on workshops.

Sort by:  

Conference:  Transform X 2022
Authors: Anima Anandkumar
2022-10-19

As the Director of AI Research at NVIDIA, Dr. Anandkumar shares some of the highest-impact topics her team is researching—from better weather modeling to tackle climate change, to modeling CO2 carbon capture and storage, to coronavirus aerosol simulation and drug discovery. Previously, Anandkumar helped launch Amazon SageMaker, Comprehend, and Rekognition at AWS (Amazon Web Services). She is also the Bren Professor at Caltech’s CMS (Computing + Mathematical Sciences) Department and serves as part of the expert network of the World Economic Forum.Building on her seminal paper that covers the initial development of tensor algorithms, Dr. Anandkumar presents her research on Fourier Neural Operators (FNOs), which in some cases can replace computationally costly Navier-Stokes equations that underpin many fluid dynamics simulations such as weather forecasting models and drug discovery processes. She advocates the use of the Fourier transform in neural networks to make them discretization- (or even quantization-) invariant. By replacing brute-force computational approaches with FNOs running on GPUs, Dr. Anandkumar is able to reduce the complexity of the simulation of weather modeling (a 45,000x speed-up) and carbon capture and sequestration (a more than 10,000x speed-up).Join Dr. Anandkumar as she will inspire attendees to apply FNOs, smarter model architectures, and parallel computation to solve the public health and climate crises of our time.
Conference:  Transform X 2022
Authors: Alan Cowen
2022-10-19

Humans evolved to communicate with other humans, not with algorithms. So, to create algorithms that are better at doing what we want, we need to understand how humans communicate. Alan Cowen, CEO and Chief Scientist at Hume AI, discusses how algorithms can better understand human communication and the role this will play in the future. Cowen covers semantic space theory, a new data-driven way of thinking about emotions and how we express them. Hume runs experiments all over the world to see how humans express themselves, and measures nuances of expression in voice, language, face, and body movements. Hume then leverages this data to fine-tune models, such as GPT-3 to create a model that controls the emotional tone of a response. The goal is to create more responsive assistive technology and to enhance training tools for healthcare professionals and others. He also discusses what new datasets and models are teaching us about the vocabulary of human expression and some of Hume’s findings, and how expressive communication and empathy can be built into modern technology. Cowen is an applied mathematician and computational emotion scientist. Prior to founding Hume AI, he was a researcher at Google AI, where he helped establish affective computing research efforts.
Conference:  Transform X 2022
Authors: Arvind Neelakantan
2022-10-19

Text embeddings are useful features in many applications including semantic search, predicting code completion, natural language, topic modeling, classification, and computing text similarity. Arvind Neelakantan, Research Lead and Manager at OpenAI, introduces the concept of embeddings, a new terminus in the OpenAI API. When OpenAI originally introduced the API two years ago, it was based on the GPT-3 model, which was useful for many tasks. But, as Neelakantan explains, GPT-3 is not explicitly optimized to produce a single vector or embedding of the input. This ability, to have a condensed representation of the input, would be helpful for programmers and others to use as features for downstream applications, the OpenAI team determined. They set about building an unsupervised model that is good at getting this kind of single embedding, and created a contrastive pre-training model, which Neelakantan will describe. He covers use cases for embeddings, and how the API is used in the real world, including at JetBrains Research for astronomical research and at FineTune Learning, which builds education systems. FineTune is using text embeddings to more accurately find textbook content based on learning objectives.
Conference:  Transform X 2022
Authors: Dragomir Anguelov, Marco Pavone, Alex Kendall, Kate Park
2022-10-19

tldr - powered by Generative AI

Experts discuss the challenges in incorporating machine learning into autonomous vehicles safely and effectively.
  • Autonomous vehicles use multiple sensors to identify their surroundings, but face difficulties in identifying pedestrians, other vehicles, obstacles, and environmental conditions.
  • Integrating complicated sensor suites, software, data management, and machine learning with engineering is a challenge.
  • Collecting and labeling large amounts of data, integrating ML models with the rest of the self-driving stack, and improving the driver continuously are also challenges.
  • Simulation plays a critical role in development.
  • Different OEMs use unique approaches to leverage machine learning in their self-driving stack, with some using end-to-end learning and others preferring modular learning.
  • Scaling to new environments quickly is a difficult challenge.
Conference:  Transform X 2022
Authors: Dr. Kenneth E. Washington, Vijay Karunamurthy
2022-10-19

tldr - powered by Generative AI

Ken Washington, CTO of Ford, discusses the responsible development of AI and its potential to benefit society in a conference presentation.
  • Developers should be responsible and focus on doing good when working with AI
  • AI has the potential to benefit society, such as in home robots that provide companionship and peace of mind
  • Ford is providing a software development kit for Astro, their home robot, to university partners to accelerate innovation
  • Working in AI is important for society and provides opportunities for learning and growth
Conference:  Transform X 2022
Authors: Daniele Perito, Jon Wilfong
2022-10-19

tldr - powered by Generative AI

The speaker reflects on the importance of making the right decisions when building data infrastructure for machine learning and the need for the field of data science to adopt software engineering principles.
  • Data infrastructure is sticky and difficult to move away from, so it's important to make the right decisions when building it
  • The field of data science and machine learning can benefit from adopting software engineering principles
  • Data scientists should be screened for coding skills and encouraged to code into the back-end code base
  • Modular, testable, and simple systems are necessary for compounding interest in machine learning work
  • The field of machine learning and AI has been heavily seeded from academia, which hasn't always used the best engineering practices
Conference:  Transform X 2022
Authors: Eric Rachlin, Natalya Tatarchuk, Vivek Muppalla
2022-10-19

tldr - powered by Generative AI

The intersection of AI, motion capture, and digital humans is critical for creating believable interactive avatars. Inference and machine learning are important for optimizing rendering pipelines and creating fast content creation. The advice for those interested in the field is to focus on shipping rather than perfection.
  • AI, motion capture, and digital humans are critical for creating believable interactive avatars
  • Inference and machine learning are important for optimizing rendering pipelines and creating fast content creation
  • Shipping is more important than perfection
Conference:  Transform X 2022
Authors: Michael I. Jordan
2022-10-19

tldr - powered by Generative AI

The speaker discusses the importance of incorporating economic principles into machine learning systems to build large-scale, economically sound structures that benefit society.
  • Rey is a platform that combines functional programming and object-oriented programming to create a model of distributed asynchronous computing.
  • The speaker believes that AI and machine learning have been missing the boat by neglecting the importance of market intelligence.
  • Building multi-way markets, exploring to learn preferences, and uncertainty quantification are some of the open academic problems in this field.
  • Fairness, privacy, and social good are important considerations that have an economic side to them.
  • The speaker provides an anecdote about UnitedMasters, a company that created a two-way market for musicians to stream their music to the NBA and get paid for it.
  • The overall goal is to build large-scale systems that are healthy, happy, and economically efficient.
Conference:  Transform X 2022
Authors: Dr. Lynne Parker, Michael Kratsios
2022-10-19

tldr - powered by Generative AI

The use of AI in the federal government can streamline processes, reform regulations, and improve citizen services. However, there is a talent challenge and a need for guidance on AI procurement.
  • AI can be used to process paperwork and summarize important information for agencies to address citizen problems
  • AI can be used for regulatory reform to detect contradictory regulations and flag them for correction
  • 13 federal agencies have made public their use cases of AI
  • There is a talent challenge in federal agencies to implement AI
  • There is a need for guidance on AI procurement to accelerate the use of AI in the federal government
Conference:  Transform X 2022
Authors: James Manyika, Alexandr Wang
2022-10-19

tldr - powered by Generative AI

The challenges of reskilling at scale and the role of humans in fine-tuning and embedding AI systems into society
  • The challenge of reskilling at scale is greater due to faster progress in technology
  • Current AI systems are intelligence machines, not learning machines like children
  • There will be millions of jobs around fine-tuning and guiding AI systems
  • Social technical embedding is necessary to effectively put AI systems into the world
  • The 'other' job category is the fastest growing and reflects new and emerging activities
  • Ethical use of AI is a concern that needs to be addressed