Upgrade to Pro — share decks privately, control downloads, hide ads and more …

AI in Action

AI in Action

AI in Action: Real-World Use Cases for Transforming Workflows and Boosting Productivity

Presented at IWD Winnipeg, Manitoba. Apr 15, 2023

Charmi Chokshi

May 26, 2023
Tweet

More Decks by Charmi Chokshi

Other Decks in Technology

Transcript

  1. Charmi, Chokshi ML Grad, Mila and UdeM AI in Action

    Real-World Use Cases for Transforming Workflows and Boosting Productivity
  2. Hey there! • ML Grad at Mila and UdeM •

    GDE for ML • International Speaker • Worked @ AWS, Shipmnts, ISRO • Tech, Art, Travel
  3. Plan of the Day! • Intro to ChatGPT • Tools

    for Coding • Image • Productivity • Business • Design • A curated list • Q&A
  4. AI is one of the most important things humanity is

    working on. It is more profound than, I dunno, electricity or fire. - Sundar Pichai
  5. Artificial Intelligence Deep Learning Machine Learning Any Technique that enables

    computers to mimic human intelligence & behaviour A subset of ML exposing multilayered neural networks to vast amount of data A subset of AI that includes statistical techniques to solve the tasks using experience
  6. ChatGPT Generative Pretrained Transformer GPTs are a family of models

    proposed by OpenAI. They can create text from scratch.
  7. ChatGPT Generative Pretrained Transformer Generative models are used to generate

    new outputs from scratch. So, given a prompt, it can imagine the rest. In the case of GPT, it can only generate text. Other models can also generate images, sounds, etc. (DALL-E, Stable Diffusion, Riffusion, ...)
  8. ChatGPT Generative Pretrained Transformer Model pretrained on *LOTS of data.

    Think ~all of the publically available text on the internet. * Models trained on LOTS of data are called Large Language Models (LLM). GPT is an example of an LLM, but there exist many others too (Flan-T5, BLOOM, ...)
  9. ChatGPT Generative Pretrained Transformer A specific architecture of neural networks.

    Transformers were first published in 2017 at NeurIPS by researchers at Google. This was a watershed moment for LLMs. It allowed researchers to train models efficiently on huge datasets. https://arxiv.org/pdf/1706.03762.pdf
  10. Prompts in ChatGPT During training, ChatGPT made billions of connections

    between trillions of words, and is able to statistically predict the next word for any ‘prompt’ (like a query or question)
  11. -VEs of ChatGPT • Writing Complex code • Arithmetic •

    Citing Sources • Being Factual • Tricky Logic • Playing some games • "Safety"
  12. Auto-GPT Based on GPT-4, an open source experimental tool that

    can come up with ‘self-prompts’ and work autonomously. Also, includes internet connectivity!
  13. Github Copilot: Your AI Pair Programmer https://github.com/features/copilot Uses OpenAI GPT-3

    to generate human-like text and is trained on publicly available code from GitHub. It is available in Visual Studio Code, JetBrains, and Neovim.
  14. Text to Video generation with Kaiber Many more to come

    https://www.kaiber.ai/ https://designs.ai/videomaker
  15. Stay tuned for… • Microsoft 365 Copilot for office products

    • Google's tool for gmail and workspace