Google I/O 2023: Bard, PaLM 2 and AI Across Google's Products

SUMMARY

In 30 seconds

Matthew Berman recaps the AI announcements at Google I/O 2023, including Bard updates, writing help in Gmail, Adobe integrations, Gemini, photo editing, Maps and PaLM 2. He is excited by the breadth of the plans but argues that Google is mostly adding AI to existing products rather than introducing a wholly new consumer experience. The video records announcements and expectations at the time, not a current feature guide.

1Bard gets new capabilitiesBerman discusses coding support, visual search and planned integrations as Google's response to growing competition in generative AI.
2Writing and image generationGmail's Help Me Write and a proposed Bard integration with Adobe Firefly illustrate AI features placed in familiar workflows.
3Multimodal modelsGemini was announced as a model in development intended to work across modalities; Berman describes its potential in terms of text, images, audio and video.
4Photos and MapsThe event showcased AI-assisted photo editing and immersive route views in Maps.
5PaLM 2 and strategyBerman describes the new model family and closes by questioning whether extending established products is sufficiently disruptive.

Takeaways

Bard's announced updates included coding, visual features and planned integrations.
Google presented generative AI features for Gmail and photo editing, and a more immersive Maps experience.
Gemini was introduced as a multimodal model under development, while PaLM 2 was presented in multiple sizes for different uses.
Berman welcomes the investment but questions whether enhancing existing products is as disruptive as a new interface.

TopicsAI Strategy Product & UX Models & training Multimodal AI

FULL TEXT

Read the edited text by chapter

Show transcript

Transcript only

Google's AI response

Berman frames the announcements against competition from Microsoft and OpenAI, then reviews Bard changes.

Google I/O 2023 brought a long list of AI announcements, from Bard updates to new features in existing Google products. I see them as part of Google's effort to respond to competition from Microsoft and OpenAI after the rise of generative AI.

Bard had not impressed me in earlier coding tests, though it had improved. Google announced changes including a dark mode, visual features, support for more than 20 programming languages and planned integrations. These are announcements from 2023, so their availability and implementation may have changed since the video.

Gmail, Firefly and Gemini

Writing assistance and image generation accompany the announcement of a multimodal model in development.

One demonstration puts writing assistance in Gmail: describe the email you need and Help Me Write drafts it using relevant context. I want to test it myself, but the demonstration suggests a way to save time on routine correspondence. Google also announced a planned Bard integration with Adobe Firefly for image generation.

Google introduced Gemini as a multimodal model under development. The goal is to handle different kinds of information, including text, images, audio and video. Berman compares this direction with other multimodal systems, while these early announcements should not be treated as proof that all capabilities were available at launch.

Photos, Maps and PaLM 2

The presentation covers editing and mapping experiences as well as a new language-model family.

Berman also mentions an Adobe augmented-reality collaboration and then turns to Google Photos. The editing demonstrations include changing lighting, removing unwanted elements and repositioning objects. The examples show how generative tools might change an existing photograph rather than only create a new image.

SLIDE 1 / 1Immersive route view in Google MapsA 3D route preview illustrates the mapping feature discussed in the video.

Google Maps' immersive route view offers a three-dimensional flyover, incorporating information such as weather and traffic. Berman compares it to an expanded Street View experience. PaLM 2, meanwhile, is presented as a multilingual model family with coding ability and several sizes, from Gecko through Unicorn, with developer access planned.

Berman's assessment

He welcomes Google's investment while questioning whether the strategy introduces a truly new product experience.

I am excited by Google's commitment to AI, but my critique is that many announcements add AI to existing products rather than present a fundamentally new experience. Google has strong reasons to extend and protect its established search business, yet that strategy leaves open the question of how disruptive its next interface will be.