4 Ways to Get Hands-On With Generative AI at ODSC East
Generative AI has the world on fire. With its applications in creativity, automation, business, advancements in NLP, and deep learning, the technology isn’t only opening new doors, but igniting the public imagination. That’s why at ODSC East, generative AI will be one of the hot topics to be discussed in our talks, workshops, tutorials, and more. Let’s take a look at what’s in store for you at ODSC East this May 9th-11th and what you’ll learn about generative AI when you attend.
Hackathon
The Generative AI hackathon is where you’ll see the latest in greatest in AI being showcased. Come and see what the experts create as they seek to create something truly unique while using the latest tools and platforms as they push the limits of generative AI to win prizes and gain recognition in the community.
Are you interested in participating? Well, during the hackathon you’ll have access to cutting-edge tools and platforms, including Weaviate and OpenAI API & ChatGPT plugins, to work on projects such as generative search and prompt engineering. Present your innovative solution to both a live audience and a panel of judges.
Keynotes
Infuse Generative AI in your apps using Azure OpenAI Service
As you know, businesses are always looking for ways to improve efficiency and reduce risk, and one way they’re achieving this is through the integration of large language models. These models are being used to generate personalized marketing text and imagery, summarize long-form reports, and create new conversation AI experiences. In this keynote, you’ll learn how with Azure OpenAI Service, businesses can leverage some of the most advanced AI models, such as Dall-E 2, GPT-3.5, Codex, and ChatGPT, to address these use cases and bring unique levels of efficiency to their operations.
You’ll also learn about Azure’s enterprise-grade capabilities, including security and privacy controls, geo-diversity, content filtering, and responsible AI. Additionally, by combining the power of ChatGPT and Azure Cognitive Search, businesses can even use conversational language for knowledge base retrieval on their enterprise data.
Unlocking the Power of Large Language Models: Why Owning Your Own Model is Critical — and Within Reach (No abstract so I created this out of the title)
The rise of large language models like GPT-3 and its successors has fundamentally transformed the field of natural language processing, enabling powerful new applications in areas like chatbots, content creation, and more. However, many businesses are hesitant to rely on these models due to concerns around ownership, data governance, data privacy, and the cost associated with integration into existing systems.
In this keynote, you’ll learn how it is now easier than ever for businesses to own and operate their own large language models, thanks to technological advances and the availability of pre-trained models from companies like MosaicML and Oracle. By owning your own model, you can unlock new levels of flexibility, control, and innovation, while ensuring that your data and intellectual property remain secure. If this piques your interest, then you won’t want to miss this keynote.
MLOps in the Era of Generative AI
ChatGPT is a fascinating example of the power and potential of Generative AI and foundation models in modern AI applications. However, it also presents a new set of operationalization challenges that must be overcome. These challenges include handling massive amounts of data, large-scale computation and memory, complex pipelines, transfer learning, extensive testing, and monitoring, among others.
In this keynote, we’ll explore the new technologies and share MLOps orchestration best practices that will enable you to automate the continuous integration and deployment (CI/CD) of foundation models and transformers, along with the application logic, in production. We’ll discuss how to manage and monitor the application pipelines at scale, and show how to use GPUs to maximize application performance while protecting your investment in AI infrastructure.
This keynote will also include a live demo of transfer learning and deployment of a transformer model using Hugging Face and MLRun; showcasing how to make the whole process efficient, effective, and collaborative. Whether you’re an experienced developer or just starting out with generative AI, this keynote will provide valuable insights into how to operationalize and deploy these powerful models in real-world applications.
Sessions
Leonardo De Marchi — Generative AI
In recent years, artificial intelligence has shown incredible potential for creativity. From generating music and poetry to creating visual art, AI models can now produce outputs that are indistinguishable from those created by humans. Leonardo De Marchi’s session will provide an overview of the different types of AI models that can be used to generate creative outputs, including NLP, computer vision, and reinforcement learning. We will take a deep dive into each of these areas, exploring how they can be used to create outputs like song lyrics, art, and designs.
But that’s not all. With Generative AI, we will also consider the ethical and legal implications of using AI to generate creative works and discuss how to ensure that these models are used responsibly and ethically.
Modern NLP: Pre-Training, Fine-tuning, Prompt Engineering, and Human Feedback
NLP is more about leveraging pre-trained, foundation models, along with fine-tuning, prompt engineering, and/or human feedback, than just creating big datasets and training models from scratch. In this workshop, participants will receive a practical, hands-on understanding of these topics, so they can start integrating modern NLP in their respective domains.
Once the workshop is over, participants will have fine-tuned and prompt-engineered state-of-the-art models like BART and XLM-Roberta. They will also get an opportunity to explore the architecture and utility of world-shaking technologies like ChatGPT, enabling them to peer behind the curtain of these innovative tools.
Topic Modeling using pre-trained large language model embeddings
When working with NLP models, traditional methods of topic modeling can fall short when dealing with small responses. This can make it difficult to find common topics in social media comments, for example, that mention a specific brand. However, using a pre-trained large language model can provide a solution. LLMs are trained on much larger datasets, which allows them to contain richer information about how words are typically used together.
By translating responses to high-dimensional vector embeddings, clustering algorithms can group the data into discrete sets of documents that share a semantic similarity. To visualize these relationships, the UMAP algorithm can reduce high-dimensional data to a 2D or 3D representation that can be easily plotted and inspected. This session will demonstrate how to find topics in brief text responses and create interactive visualizations of the results using free open-source Python packages.
A Zero-shot 2D Sentiment Model Predicts Clinical Outcome in Psilocybin
Over the years, the therapeutic use of psychedelic drugs has been shown to have the potential in improving mental health as seen in both historical accounts and scientific studies. A recent study demonstrated the safety and effectiveness of psilocybin in treating patients with treatment-resistant depression. However, the treatment does not work for all patients and unintended effects are always a major worry, so predicting the outcome is crucial. Researchers used a zero-shot machine learning classifier based on the BART model to predict the treatment outcome using transcripts from psychological support sessions.
The algorithm accurately predicted long-term patient response, allowing for personalized treatment and optimization of the therapeutic model. This session will discuss the algorithm and how further research using models can help to predict treatment outcomes.
Learning from Jeff Clune Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos
Join Jeff Clune, an Associate Professor of Computer Science at the University of British Columbia, as speaks about an exciting new paradigm in machine learning that extends the GPT model to learning to act by watching videos. In this session, he’ll showcase how using a massive dataset of unlabeled human Minecraft play videos, he and his team were able to train a neural network to play Minecraft using only a small amount of labeled contractor data.
The result was a model was able to learn complex tasks, such as crafting diamond tools, in a fraction of the time it would take a proficient human player. This was done by using the native human interface of keypresses and mouse movements to teach the model. The model is highly general and represents a step towards creating more intelligent computer-using agents. Clune and his team believe that this paradigm could help automate much of the work currently done by humans on computers, with far-reaching implications for industries across the board, especially administrative and repetitive tasks.
ODSC East AI Expo & Startup Showcase
In addition to the insightful sessions and workshops at ODSC East, attendees will also have the opportunity to explore the AI Expo & Startup Showcase. This is the perfect place to witness the latest developments and advancements in the field of generative AI and beyond. Here, leading startups will be present to showcase their cutting-edge products, including fascinating demos, presentations, & more.
Whether you’re an industry professional, researcher, or simply interested in the latest trends in AI, the AI Expo is an excellent opportunity to network and engage with like-minded individuals. The AI Expo & Startup Showcase promises to be an exciting and enlightening experience, with a diverse range of exhibits and talks to enjoy. So be sure to make some time in between workshops, keynotes, and other events to get the most out of the AI Expo at ODSC East.
Conclusion
Some neat stuff right? ODSC East has everything that you’re looking for when it comes to generative AI. From Keynotes, talks, hands-on workshops, tutorials, and the AI Expo & Startup Showcase, you’ll experience the entire spectrum of generative AI from the leaders who are pioneering the technology.
If you register now, you’ll get 30% off your ticket. But you have to register now as tickets are running out fast!
Originally posted on OpenDataScience.com
Read more data science articles on OpenDataScience.com, including tutorials and guides from beginner to advanced levels! Subscribe to our weekly newsletter here and receive the latest news every Thursday. You can also get data science training on-demand wherever you are with our Ai+ Training platform. Subscribe to our fast-growing Medium Publication too, the ODSC Journal, and inquire about becoming a writer.
