AI for everyone with Gemma
Skills:
AI Pair Programming80%
Key Takeaways
Empowers developers to build the future with Gemma, a family of powerful open models, for community-driven innovation
Full Transcript
[Music] hi everyone my name is Kathleen canel I'm a Staff research engineer at Google deepmind and one of the technical leads of the Gemma team before I get started I just want to say what an honor it is to be part of the women in AI Summit now more than ever it's so important and impactful for us to gather as a community and I'm proud to say that Gemma is a team where more than 50% of our leads are women so I'm incredibly proud to be here representing the women of Gemma and the Gemma team at Google deepmind we dream of a world where AI unlocks solutions to Humanity's grandest challenges we've been pushing the frontiers of what's possible Alpha fold revolutionized protein folding and recently won the Nobel Prize Alpha zero mastered go and chess and now we have Gemini Gemini is our most ambitious project yet a multimodal AI fluent in text images audio and code it reasons it understands it creates it's a glimpse into a world where the limits of human Ingenuity are Amplified by the power of intelligent machines but raw power is not enough true progress lies in Sharing access to these transformative tools with developers like you and that's why we built Gemma built from the same Cutting Edge research and Technology as Gemini Gemma is a family of lightweight open models meticulously crafted for you we want to put this power in your hands we want to see what you can build we launched Gemma just 8 months ago and we've been absolutely Blown Away by the response a Vibrant Community which we like to call the gem verse has has sprung up around it creating an incredible ecosystem of both model variants and tools in fact there are already over 40,000 Gemma variants published on hugging face this enthusiastic adoption fueled us to accelerate our development and just a few months later we launched Gemma 2 Gemma 2 comes in three sizes a 2 billion parameter Model A a 9 billion parameter model and a 27 billion parameter model each designed with a specific purpose in mind the 27 billion Pam model is our Powerhouse shortly after it was launched the 27b model stormed the LM Cy chatbot Arena leaderboard outperforming models to to 10 times its size but what's truly remarkable is the performance we achieved even with smaller models our 2 billion parameter model actually outperforms GPT 3.5 on LMS a model many many times its size just imagine the possibilities being able to access this level of performance offline anytime anywhere this is the future of AI powerful portable and personal and for those looking for a balance between power and efficiency the 9 billion parameter model has proven incredibly popular and performant and importantly this performance is available to everyone at the heart of the Gemma team is a commitment to building amazing models for real people for you we design these models to fit on your Hardware the hardware developers already have access to you don't need giant clusters of gpus or tpus to harness the power of demo and crucially we refuse to sacrifice quality our goal is to deliver cuttingedge performance in a package that's practical accessible and empowers you to build incredible things but Gemma is more than just benchmarks it's about responsible AI prioritizing safety and accessibility that's why we recently launched Shield Gemma our suite of safety classifiers and gemos scope a tool that enables unprecedented model interpretability Gemma is less than 8 months old and we've been incredibly inspired by what the Gemma Community has already achieved we're particularly excited about stories like this one about a group of developers in India let's take a [Music] look language is a is a very interesting problem to solve actually and given India has a huge variety of languages and it changes every 5 kilm when technology is developed for a particular culture it won't be able to solve and understand the nuances of a country like India one of gemma's features is an incredibly powerful tokenizer which enables the model to use hundreds of thousands of words symbols and characters across so many alphabets and language systems this large vocabulary is critical to adapting Gemma to power projects like navasa navasa is a model that's trained for indic languages it's a fine tun model based on Google's Jama we built narasa to make large language models culturally rooted where people can talk in their native language and get the responses in their native language our biggest dream is to build a model to include everyone from all corners of India today the language that you speak in could be the tool and the technology that you use for solving your real world problems and that's the power of generative AI that we want to bring to every corner of India and the entire world we're constantly Amazed by stories like this we've seen remarkable work particularly in adapting Gemma for languages and tackling regionally specific challenges you've published Gemma models for underrepresented languages like Arabic Ukrainian Zulu and many more Beyond national languages we're also seeing efforts to preserve and promote Regional dialects in Korea for instance a developer is leveraging Gemma to build a dialect translator specifically focused on the G dialect juu Island a beautiful volcanic island off Korea's southern coast boasts a unique language distinct from standard Korean this is exactly the kind of progress that fuels our passion we want to see a world where everyone has access to helpful AI no matter where they are or which device they own so we've also released our tuning recipes empowering you to create Gemma models for your language your culture and to accelerate this Vision we've also announced a $100,000 prize towards our newest kaggle competition to build a Gemma model for your language and share your knowledge with the world and compete for prices let's build a truly Global AI ecosystem together these are just a few of the applications for Gemma that we are particularly excited about because we know that real progress will come from you your achievements continue to push the boundaries of what is possible and it is a true honor for us to be a part of your journey the future is in your hands let's build it together thank you [Music]
Original Description
Explore how Gemma, a family of powerful open models, is empowering developers to build the future. Discover community-driven innovation and the exciting potential of AI to solve humanity's great challenges - regardless of hardware limitations or geographic location.
Resources:
Gemma 2 on Kaggle → https://goo.gle/3B1uWeZ
Gemma 2 on Hugging Face → https://goo.gle/4fSNjBO
Gemma Cookbook → https://goo.gle/gemma-cookbook
Competition: Unlock Global Communication Kaggle → https://goo.gle/language-ai
Join the Build with Google AI developer forum → https://goo.gle/3OHNVy4
Check out more Women in AI Summit 2024 sessions → https://goo.gle/WiAI24-all
Subscribe to Google for Developers → https://goo.gle/developers
#AI
Speaker: Kathleen Kenealy
Products Mentioned: Gemma
Event: Women in AI Summit 2024
Watch on YouTube ↗
(saves to browser)
Sign in to unlock AI tutor explanation · ⚡30
Playlist
Uploads from Google for Developers · Google for Developers · 0 of 60
← Previous
Next →
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Developer Journey - Sunnyvale DSC Summit ‘19
Google for Developers
How Google is working with students - Sunnyvale DSC Summit ‘19
Google for Developers
Starting your career in the Cloud - Sunnyvale DSC Summit ‘19
Google for Developers
The Solution Challenge - Sunnyvale DSC Summit ‘19
Google for Developers
Firebase - Sunnyvale DSC Summit ‘19
Google for Developers
Cloud Hero - Sunnyvale DSC Summit ‘19
Google for Developers
Panel discussion - Sunnyvale DSC Summit ‘19
Google for Developers
The art of negotiation - Sunnyvale DSC Summit ‘19
Google for Developers
Courage to care, solve and share - Sunnyvale DSC Summit ‘19
Google for Developers
Version 9 of Angular, Glass Enterprise Edition 2, path to DX deprecation, & more!
Google for Developers
[DEPRECATING] Introducing a new series (Assistant for Developers Pro Tips)
Google for Developers
Detecting memory bugs with HWASan, Bazel 2.1, Next ‘20 session guide, & more!
Google for Developers
Why Podcast.app chose a .app domain name
Google for Developers
Machine Learning Bootcamp Jakarta 2019
Google for Developers
Android Studio 3.6, Android 11 Developer Preview, Kubeflow 1.0, & more!
Google for Developers
[DEPRECATING] Importance of community (Assistant on Air)
Google for Developers
Why the Flutter team switched from .io to a .dev domain name
Google for Developers
3 website-building tips from .dev creators
Google for Developers
Why NimbleDroid chose a .app domain name
Google for Developers
Android Platform Codelab, Bazel 2.2, Maps Android Utility Library v1.0, & more!
Google for Developers
Google for Games Developer Summit: A free, digital experience for game developers
Google for Developers
Inspecting Home Graph (Assistant for Developers Pro Tips)
Google for Developers
Google for Games Developer Summit Keynote
Google for Developers
Stadia Games & Entertainment presents: Keys to a great game pitch (Google Games Dev Summit)
Google for Developers
Empowering game developers with Stadia R&D (Google Games Dev Summit)
Google for Developers
Supercharging discoverability with Stadia (Google Games Dev Summit)
Google for Developers
Stadia Games & Entertainment presents: Creating for content creators (Google Games Dev Summit)
Google for Developers
Bringing Destiny to Stadia: A postmortem (Google Games Dev Summit)
Google for Developers
Live Captioning in Google Slides
Google for Developers
[DEPRECATING] User engagement for the Google Assistant
Google for Developers
TensorFlow Dev Summit ‘20, Google for Games Dev Summit, Cloud AI Platform Pipelines, & much more!
Google for Developers
Top 5 from the TensorFlow Dev Summit 2020
Google for Developers
Developer Student Clubs 2019 Turkey Leads Summit
Google for Developers
Building simpler payment experiences | Google Pay Plugin for Magento 2
Google for Developers
Become A Developer Student Club Lead
Google for Developers
Firebase Kotlin Extensions, ARM apps on the Android Emulator, Angular v9.1, & more!
Google for Developers
Test suite for Smart Home (Assistant for Developers Pro Tips)
Google for Developers
Google Play updates, Bazel 3.0, Business Console for Google Pay, & more!
Google for Developers
How to use error logs (Assistant for Developers Pro Tips)
Google for Developers
Contact Center AI, Android Studio 4.1 Canary 5, TensorFlow QAT API, & more!
Google for Developers
WebView DevTools, Kotlin meets gRPC, Flutter CodePen support, & more! (Episode 200)
Google for Developers
Offline handling for Smart Home (Assistant for Developers Pro Tips)
Google for Developers
Android 11 Dev Preview 3, Google Fonts for Flutter, Shielded VM, & more!
Google for Developers
Machine Learning Foundations: Ep #1 - What is ML?
Google for Developers
Flutter web support updates, BigQuery materialized views, Cloud Spanner emulator, & more!
Google for Developers
Computer vision by building a neural network with TensorFlow | Machine Learning Foundations
Google for Developers
Machine Learning Foundations: Ep #3 - Convolutions and pooling
Google for Developers
Android 11 Beta plans, Flutter 1.17, Dart 2.8, & much more!
Google for Developers
Machine Learning Foundations: Ep #4 - Coding with Convolutional Neural Networks
Google for Developers
Google Developers ML Summit
Google for Developers
Real-world image classification using convolutional neural networks | Machine Learning Foundations
Google for Developers
Adobe XD support for Flutter, Architecture Framework, temporary closures with Places API, & more!
Google for Developers
Machine Learning Foundations: Ep #6 - Convolutional cats and dogs
Google for Developers
Machine Learning Foundations: Ep #7 - Image augmentation and overfitting
Google for Developers
Announcing Firebase Live, Flutter Day, Java 11 on Google Cloud Functions, & more!
Google for Developers
Machine Learning Foundations: Ep #8 - Tokenization for Natural Language Processing
Google for Developers
Android 11 Beta, Google Play Asset Delivery, Firebase Crashlytics SDK, & much more!
Google for Developers
Natural Language Processing: Using sequencing APIs in TensorFlow | Machine Learning Foundations
Google for Developers
Build a sarcasm classifier using NLP and TensorFlow | Machine Learning Foundations
Google for Developers
AR Realism with the ARCore Depth API
Google for Developers
More on: AI Pair Programming
View skill →Related Reads
📰
📰
📰
📰
One Brain, Two Wallets: Two Claude Code Accounts on One Machine
Dev.to AI
We Built an AI Assistant for Word That Actually Formats Your Documents (And Runs Locally)
Dev.to AI
How I Built an AI Interview Prep App with Spring Boot + RAG in 10 Days
Dev.to · sandeep gusain
Your AI Re-Reads Your Whole Codebase Every Session. Hand It a Map Instead
Dev.to · Athreya aka Maneshwar
🎓
Tutor Explanation
DeepCamp AI