Generative AI
The future of technology is here,
and it’s generative!
Toby Vincent John
3rd Year B.Tech AI/ML student at VIT-Chennai
Passionate about solving real-world problems
Exploring AI, ML and full stack development
Interests - Hackathons & Gaming
Winner - 2024 Google Gen AI Exchange Hackathon
Agenda
Introduction to Generative AI
Boom of Generative AI
The current state of AI
Career Paths in Generative AI
The Build Fast Ship Fast culture
Networking for AI
Future of GEN-AI
Q & A
Introduction to
Generative AI
GENERTIVE AI REFERS TO A SUBSET OF ARTIFICIAL INTELLIGENCE
THAT FOCUSES ON CREATING NEW CONTENT, SUCH AS TEXT,
IMAGES, VIDEOS AND OTHER MEDIA BASED ON USER PROMPTS.
Pre-2017: Early neural networks, RNNs, Word2Vec, GANs.
2017: Transformer model introduced - “Attention is all you need”
2018: BERT revolutionizes NLP with bidirectional context.
2019: GPT-2 showcases large-scale unsupervised generation.
2020: GPT-3 brings vast improvements in generative text capabilities.
2021: Multimodal models (CLIP, DALL·E).
2022: Stable Diffusion launches opensource text-to-image generation.
2023: GPT-4, ChatGPT, and ethical concerns.
2024: Continued evolution with multimodal models and practical applications.
The Boom of Generative AI
Important Milestones
What people think AI is about?
Just chatbots
Only for Image Generation
This misconception isn’t just limited to the common
man, but also to a majority of AI “enthusiasts”
“It’s too boring. All that people
build are GPT wrappers that answer
questions that every AI can answer”
Autonomous AI agents - Game Characters, Marketing Agents, Database Agents
AI pipelines for automation
Music and Sound generation - OpenAI’s Jukebox, Google’s MusicLM, Suno, Udio
Video Generation and Editing: RunwayML, Hailuo, Kling
Code Generation & Development: Generate code, automate coding tasks,
refactoring, bug fixes, test case generation
Scientific Discovery: Drug discovery and molecular design - DALL E for
molecules, DeepChem
What generative AI can really do
How to start a career in Generative AI
Two main paths to choose
AI/ML Researcher - New algorithms, improve
existing models
AI Developer - Real life solutions and deploying
models for practical applications
These paths have a certain degree of overlap but they’re
the 2 main groups of people working with AI
How do i pursue research?
Main focus - New Algorithms, Improving model
architecture and accuracy
Main subjects needed - Math, probability and
statistics
Skills needed Python, Pytorch, Tensorflow
Journal and Conference Publications - JAIR,
Springer, Elsevier, NeurIPS, ICML, ICLR
Degree - Preferably a Ph.D in Machine Learning or
Artificial Intelligence
Future as a research scientist at Google AI,
Microsoft Research, Meta AI, OpenAI, etc.
Main focus: Build practical AI Applications
and deploy models for real world use cases
Skills Needed:
Strong Python fundamentals, knowledge of
open source frameworks like llama index and
langchain
Ability to work with pre-trained models and
frameworks like ChatGPT, Stable Diffusion,
Hugging Face
Cloud computing platforms - GCP and Vertex
AI, AWS, Microsoft Azure
How do I become an AI developer?
Build Fast Ship Fast with AI
Stack - NextJS (Frontend), FastAPI (Backend), Docker for containerisation, AI tools like Github Copilot, CursorAI
HuggingFace for opensource models/ inference API’s for testing AI
Platforms like Google Cloud Platform and huggingface to make training and workflows easier
Where can i get started now: Create a GCP account to avail $300 of credit to get started with building
My advice: Don’t build it as projects over the summer, build it as something you would intend to sell with a use case that
you believe in, and learn your stack along the way, or to start by diving into hackathons.
Pitfalls: Dissatisfaction just working with GEN AI and not seeing any progress at building a product or not being able to
make any real change.
Solution: Build with like minded teammates to cover for frontend / backend
Progress in the order of GEN-AI->Backend->Frontend if you want to build solo
My hackathon journey
Google GenAI Exchange Hackathon
October 2nd - 18th, 2024
Cybersecurity Audit Analysis
RAG, Knowledge Graphs, Agents
Tech Stack - NextJS, FastAPI,
Challenges
Solution - CyberStrike AI
Networking
Put yourself out there
Hackathons - Devfolio, Hack2Skill, HackerEarth, lablab.ai, Unstop, Kaggle
Hackathons going on currently? - Google AI for Impact, Amazon Smbhav hackathon and
countless others in these platforms
Want to go beyond just hackathons and make an actual difference but need capital?
Apply for Grants/funding - aigrant.com, ycombinator.com, win at lablab.ai
Huggingface Blogs/ Opensource with github to meet people with experience to help you
and also refer to OpenAI blogs.
Try getting into AI startups you learn a lot - Embrace the Build Fast Ship Fast culture
Future of GEN-AI
AGI and True Intelligence
What is intelligence?
AGI and its impact - Will AI replace Software engineers?
True Intelligence - Will we ever have Jarvis like AI?
Energy based models like JEPA to replace LLMs?
Biggest contenders in this field
Too early to win, too late to know
Sir Demis Hassabis
Co-Founder - Deepmind (Google)
Yann LeCun
VP - Meta AI
Q & A
Thank You

Generative AI presentation by GOOGLE DEVELOPER GROUPS

  • 1.
    Generative AI The futureof technology is here, and it’s generative!
  • 2.
    Toby Vincent John 3rdYear B.Tech AI/ML student at VIT-Chennai Passionate about solving real-world problems Exploring AI, ML and full stack development Interests - Hackathons & Gaming Winner - 2024 Google Gen AI Exchange Hackathon
  • 3.
    Agenda Introduction to GenerativeAI Boom of Generative AI The current state of AI Career Paths in Generative AI The Build Fast Ship Fast culture Networking for AI Future of GEN-AI Q & A
  • 4.
    Introduction to Generative AI GENERTIVEAI REFERS TO A SUBSET OF ARTIFICIAL INTELLIGENCE THAT FOCUSES ON CREATING NEW CONTENT, SUCH AS TEXT, IMAGES, VIDEOS AND OTHER MEDIA BASED ON USER PROMPTS.
  • 5.
    Pre-2017: Early neuralnetworks, RNNs, Word2Vec, GANs. 2017: Transformer model introduced - “Attention is all you need” 2018: BERT revolutionizes NLP with bidirectional context. 2019: GPT-2 showcases large-scale unsupervised generation. 2020: GPT-3 brings vast improvements in generative text capabilities. 2021: Multimodal models (CLIP, DALL·E). 2022: Stable Diffusion launches opensource text-to-image generation. 2023: GPT-4, ChatGPT, and ethical concerns. 2024: Continued evolution with multimodal models and practical applications. The Boom of Generative AI Important Milestones
  • 6.
    What people thinkAI is about? Just chatbots Only for Image Generation This misconception isn’t just limited to the common man, but also to a majority of AI “enthusiasts” “It’s too boring. All that people build are GPT wrappers that answer questions that every AI can answer”
  • 7.
    Autonomous AI agents- Game Characters, Marketing Agents, Database Agents AI pipelines for automation Music and Sound generation - OpenAI’s Jukebox, Google’s MusicLM, Suno, Udio Video Generation and Editing: RunwayML, Hailuo, Kling Code Generation & Development: Generate code, automate coding tasks, refactoring, bug fixes, test case generation Scientific Discovery: Drug discovery and molecular design - DALL E for molecules, DeepChem What generative AI can really do
  • 8.
    How to starta career in Generative AI Two main paths to choose AI/ML Researcher - New algorithms, improve existing models AI Developer - Real life solutions and deploying models for practical applications These paths have a certain degree of overlap but they’re the 2 main groups of people working with AI
  • 9.
    How do ipursue research? Main focus - New Algorithms, Improving model architecture and accuracy Main subjects needed - Math, probability and statistics Skills needed Python, Pytorch, Tensorflow Journal and Conference Publications - JAIR, Springer, Elsevier, NeurIPS, ICML, ICLR Degree - Preferably a Ph.D in Machine Learning or Artificial Intelligence Future as a research scientist at Google AI, Microsoft Research, Meta AI, OpenAI, etc.
  • 10.
    Main focus: Buildpractical AI Applications and deploy models for real world use cases Skills Needed: Strong Python fundamentals, knowledge of open source frameworks like llama index and langchain Ability to work with pre-trained models and frameworks like ChatGPT, Stable Diffusion, Hugging Face Cloud computing platforms - GCP and Vertex AI, AWS, Microsoft Azure How do I become an AI developer?
  • 11.
    Build Fast ShipFast with AI Stack - NextJS (Frontend), FastAPI (Backend), Docker for containerisation, AI tools like Github Copilot, CursorAI HuggingFace for opensource models/ inference API’s for testing AI Platforms like Google Cloud Platform and huggingface to make training and workflows easier Where can i get started now: Create a GCP account to avail $300 of credit to get started with building My advice: Don’t build it as projects over the summer, build it as something you would intend to sell with a use case that you believe in, and learn your stack along the way, or to start by diving into hackathons. Pitfalls: Dissatisfaction just working with GEN AI and not seeing any progress at building a product or not being able to make any real change. Solution: Build with like minded teammates to cover for frontend / backend Progress in the order of GEN-AI->Backend->Frontend if you want to build solo
  • 12.
    My hackathon journey GoogleGenAI Exchange Hackathon October 2nd - 18th, 2024 Cybersecurity Audit Analysis RAG, Knowledge Graphs, Agents Tech Stack - NextJS, FastAPI, Challenges Solution - CyberStrike AI
  • 13.
    Networking Put yourself outthere Hackathons - Devfolio, Hack2Skill, HackerEarth, lablab.ai, Unstop, Kaggle Hackathons going on currently? - Google AI for Impact, Amazon Smbhav hackathon and countless others in these platforms Want to go beyond just hackathons and make an actual difference but need capital? Apply for Grants/funding - aigrant.com, ycombinator.com, win at lablab.ai Huggingface Blogs/ Opensource with github to meet people with experience to help you and also refer to OpenAI blogs. Try getting into AI startups you learn a lot - Embrace the Build Fast Ship Fast culture
  • 14.
    Future of GEN-AI AGIand True Intelligence What is intelligence? AGI and its impact - Will AI replace Software engineers? True Intelligence - Will we ever have Jarvis like AI? Energy based models like JEPA to replace LLMs? Biggest contenders in this field Too early to win, too late to know Sir Demis Hassabis Co-Founder - Deepmind (Google) Yann LeCun VP - Meta AI
  • 15.
  • 16.