Google I/O 2023, like past editions of the event, has made several exciting announcements that have the potential to change the way we live, particularly with Generative AI.
Google’s annual developer conference is always a showcase for the company’s latest and greatest technologies. At the event, the company provides valuable insights into future directions, as well as tools and resources to help create innovative applications, services and solutions.
There were hundreds of launches at Google I/O 2023, but in my opinion the top 5 are listed below. Check out!
1. PaLM 2
PaLM 2 is a large language model (LLM) trained on a huge dataset of text and code. According to Google, this model has been trained in over 200 languages and 20 programming languages. The model is capable of performing a variety of tasks, including advanced thinking tasks, code and math, sorting and answering questions, as well as language translation.
PaLM 2 is already being used to power Bard’s expertise, such as support for over 20 programming languages, including Python, JavaScript, Fortran, C++, Verilog, and Prolog. As a result, Bard can now write, debug and modify code in different programming languages.
The interesting about PaLM 2 is that it has four different model sizes, and a spotlight was given to Gecko, which is so lightweight that it can be used to create mobile devices applications, even offline.
Finally, PaLM 2 is also being used in other next-generation models such as Med-PaLM 2 (for health data) and Sec-PaLM (for cybersecurity data).
Med-PaLM 2 is an LLM specializing in medical knowledge. This version is trained and equipped with medical texts, including journals, textbooks, clinical notes, and patient records. According to Google, Med-PaLM 2 can provide answers to medical exam problems at a physician’s expert level with an accuracy level of 85%.
So far, the company is opening up Med-PaLM 2 to a closed group of Google Cloud customers for testing and feedback.
Sec-PaLM 2 is designed to detect security threats. It can analyze and explain the behavior of potentially malicious scripts and detect which are really threats to people and organizations. The model was trained on many publicly available security datasets such as malware samples, vulnerability reports, exploit code, phishing emails, etc.
With this, it can understand and generate security-related texts and codes in different languages and formats. Will we see Sec-PaLM doing the work of current SAST/DAST tools, or will we see integration and synergy between them?
2. Vertex AI
Vertex AI, machine Learning (ML) platform, was one of the most important announcements at Google I/O 2023 in my opinion. The tool lets you train and deploy ML models and AI applications by providing a variety of tools and services that make getting started with ML easy, even for developers with no experience in the subject.
In fact, Vertex AI is replacing the old Google AutoML Vision platform and, on top of that, expanding with AI services. An example was the announcement of Generative AI Studio, which makes it easy to rapidly prototype and test generative AI models, with support for natural language chat.
We also had the announcement of three new Generative AI models that are available in Vertex AI: Codey (text to code), Imagen (text to image) and Chirp (audio to text). Making a parallel with Open AI models, Codey is similar to Codex used by Github Copilot, Imagen is similar to DALLE and Chirp is similar to Whisper.
3. Duet AI
This was one of my favorites! Duet AI is a new collaborator with Artificial Intelligence and has been integrated into several Google systems, from the entire Workspace platform, Google Cloud and AppSheet, for example.
Duet AI is available in all major Google Workspace apps, including Gmail, Docs, Sheets, Slides, and Meet. With the use of Generative AI we can create content or structure documents with just a few prompts.
For example, Google showed how it could write a detailed email and create a spreadsheet when asked for. To test these features, you can join the Google Labs waitlist.
On Google Cloud, Duet AI can help with various tasks such as code assistance, for example, creating functions that will run in the cloud, in addition to helping with all infrastructure provisioning and deployment.
It is still possible to chat with Duet AI in a chat interface and ask specific questions about the cloud.
Finally, Duet AI is also powering AppSheet (Google’s low-code development platform). With this, AppSheet gains new capabilities and will allow users to build intelligent business applications, connect their data and create workflows in Google Workspace through natural language.
With no coding required, users will be able to build applications by describing requirements in an AI-powered prompt-guided chat. Looks promising!
4. Gen App Builder
Gen App Builder (short for Generative AI App Builder) is a tool that can help you quickly and easily build enterprise applications that leverage Generative AI.
It can be used to create chatbots, assistants, custom search engines, and all using private knowledge bases. Its interface is very intuitive and simple and promises!
5. What’s new in the mobile and Android ecosystem
There was so much news that this will be a summary.
We start with the announcement of Android 14 which is coming out at the end of the year. The new generation of smartphones in the Pixel line, the Pixel 7a, Pixel Fold, and also several updates to Android Auto, Wear, TV, etc.
For app developers, we’ve had the launch of Android Studio Bot, a new AI-powered assistant that can help developers with a variety of tasks in Android Studio, from code generation and troubleshooting to guiding through best practices, chat, and much more.
On Google Play we have also seen the use of Generative AI. The developers can use the assistant to write descriptions for the app, this is still in the experimental phase and English only, but it is interesting.
Firebase as usual has received so many updates that it’s worth a separate post, but we can highlight the new Python support when writing functions to facilitate AI adoption and several new extensions. Among them, is one that allows you to create a simple chatbot in the app with the PaLM API.
Other highlights include the addition of Terraform support for feature provisioning, Remote Config with live updates for using Feature Flags, and more security features for apps with App Check.
Also, you should check out the Android UI Design Hub, a guide to creating Android apps with a modern and consistent design, regardless of device, be it your smartphone, tablet, TV, or smartwatch.
In summary, Google I/O 2023 showcased a host of exciting features and announcements, especially highlighting the use of Generative Artificial Intelligence such as PaLM 2, Vertex AI, Duet AI, and Gen App Builder, among others.
If you’re curious about the other solutions announced by Google, check out the list of the top 100 launches made at the event.
Ricardo Lecheta, mobile specialist and Chief Technology Officer at BRQ
