Google I/O 2023 Keynote Recap: Tunnel Vision on Generative AI
What marketers need to know from Google’s annual developer conference about its grand plan to integrate generative AI across its core products
On Wednesday, Google CEO Sundar Pichai opened the 2023 Google I/O keynote with a promise that the company is on a mission to “make AI helpful for everyone” with a “bold and responsible approach.” Acknowledging the fact that AI, especially generative AI, is now at an inflection point for technological breakthrough and mainstream adoption, Google is feeling the pressure from rising competition, led by Microsoft and OpenAI, and shows its determination to accelerate the productization of its AI prowess to maintain parity with competitors.
Expectedly, this year’s I/O keynote is all about how Google plans to incorporate generative AI, powered by its latest PaLM 2 large language model (LLM), into its wide array of consumer-facing and enterprise products. Considering that “six of Google’s core services, such as YouTube, Gmail, and Google Calendar, each serve more than 2 billion users each month,” generative AI is about to get an even bigger push into the mainstream, potentially upending consumer behaviors in search and discovery and opening new channels for brands to explore.
Let’s look at some of the key areas that brand marketers should keep an eye on.
Bard Unleashed: Accelerating the AI Search Transformation
Globally, Google still accounts for over 92% of the global search market share, per the latest tracking data from Stat Counter. Yet, with Microsoft aggressively integrating OpenAI’s ChatGPT into Bing search and drumming up buzz for being first-to-market with AI search, Google is understandably feeling the pressure and fastening its pace with the AI search rollout.
First up, no more waitlist for Bard! Starting on Wednesday, Google is making its AI-powered chat mode for Google search available to everyone. Furthermore, it also got an under-the-hood upgrade with PaLM 2, the aforementioned LLM that Google opened to preview, which promises stronger capabilities for popular tasks such as writing code, analyzing images, and translating.
One of the main ways that billions of consumers will soon encounter Bard is through the “snapshots” that Google will put at the top of search results, with links to sources that corroborate the answers compiled by AI. The prominent feature in Google’s search interface, which has not changed much over the past decade or so, is a major push to familiarize mainstream users with AI-generated search results.
Of course, Bard will still have its dedicated user interface for more in-depth conversations. The interface, which Google calls “the conversational mode of Google search,” is also getting a major revamp, with integrations of images pulled from Google’s Knowledge Graph, visual search powered by Google Lens, multi-language support starting with Japanese and Korean, and more. One notable call-out here for brands is the ability to display local search results generated by Bard on Google Map, which may open further opportunities in local discovery down the road.
Keeping up with OpenAI’s ChatGPT plugins, Google also unveiled third-party plug-ins for Bard, with the usual slate of early adopters (including the likes of Instacart, Kayak, Walmart, Spotify, and Zillow) signing on to explore using Bard to reach search audiences and surface their products in a new way.
With this rollout, Google officially kicked off the transition from the Search Engine model to a Personal Assistant model, and with this change comes more mediated interactions with online content. What this changes in consumer behavior compared to the ten-blue-links search results is that people can now have a natural conversation with their search engine and ask follow-up questions or clarifications. For example, if you want to buy a new car and you ask Bard about the pros and cons of buying an electric model, it will give you a detailed answer with both advantages and disadvantages, rather than sending you to a bunch of websites with potentially conflicting information. In other words, the answer you get from Bard is already synthesized from a variety of sources, which means that if you trust Bard, there will be less incentive for you to click through to the source websites.
That said, Google does acknowledge that “people will always value the input of other people, and a thriving web is essential to that.” It will be interesting to see what upcoming initiatives it will roll out to support online publishers and media partners to compensate for potential lost revenues due to this AI-fueled intermediation.
For brand marketers, this splashy integration of Bard into Google’s main search result will also have a major impact on product discovery and corresponding SEO tactics. Google announced that as part of the integration, the Bard answers will contain product information pulled from its Shopping Graph, such as images, prices, and links to retailer websites, in response to relevant search queries. Google says its Shopping Graph is updated 1.8 billion times every hour, which connects billions of products from thousands of merchants across the web. Notably, the “anti-Amazon” alliance most definitely stands to benefit from this integration, as most of them are big box retailers that have extensively indexed their inventories for Google Shopping Graph.
Google briefly mentioned during the keynote that it is exploring ways to incorporate sponsored content and products into Bard’s answers, but stopped short of providing more details on that. Frustrating as that may be for marketers and retailers eager to find out how to retool their SEO strategies, the truth is that this is still an early-stage development, and no one, including Google and Microsoft, has really figured out how to incorporate sponsored content into AI-generated answers without compromising the quality and relevance of the search results. Similar challenges have long existed within voice search, and while text-based search should provide more digital real estate to experiment with, similar challenges remain in closing the Ai trust gap. For now, brand marketers will have to wait and see how the implementation of sponsored content in AI search plays out.
Duet AI: Boosting Productivity & Empowering SMBs
Besides search, Google’s plan to bring generative AI to its billions of users also includes a big push of features powered by PaLM 2 into its popular services productivity tools. With generative AI features designed to enhance productivity and creativity becoming an increasingly common use case for enterprise users and digital creators, Google’s keynote announcements indicate that this is another arena in which it would like to give Microsoft, the dominant enterprise software player, a run for its money.
First up, as an answer to Microsoft’s AI Copilot features rolling out across Microsoft 365 apps and services, Google is rebranding its AI features that have trickled into Google Workspace (which includes Google Docs, Sheets, Slides, and more) under “Duet AI for Workspace,” which promises to help hundreds of millions of users drafting emails, documents, slides, and more, by generating suggestions based on the context and user preferences. Duet AI can also create custom images for slides using generative AI models, or even generate natural-sounding speaker notes based on the information on the slides.
As part of Duet AI, a new “Help me write” feature, which seems to be a renamed and improved version of the existing Smart Compose feature in Gmail, will assist users by offering prose suggestions for use in Google Docs, Slides, Sheets, and Gmail. During the keynote, Google demonstrated how to leverage Duet AI to generate detailed job descriptions based on some simple input such as job titles and salary levels.
The democratization of coding is another theme that emerged during this I/O keynote. The enhanced coding capability of PaLM 2 will help more developers to write code in natural language via Bard, which now supports 20+ coding languages. Users can simply type in what they want to code in natural language, and Bard will generate the corresponding code snippets, complete with code citation and code improvement features to help users find relevant sources and optimize their code quality. It is worth noting that Microsoft also recently invested in a British startup called Builder.ai and plans to integrate its AI assistant Natasha that helps non-coders build apps into Microsoft Teams stores.
Another cool feature with enterprise use cases that Google is testing is a new translation service called “Universal Translator,” which re-dubs video in a different language while also synchronizing the speaker’s lips with words they never spoke. For brands, especially with global presence, there is a clear opportunity to leverage it to create localized video assets for dynamic targeting or internal training programs.
B2B brands and marketers should not underestimate the productivity boost that generative AI promises to bring to their clients, especially when it comes to empowering small and medium businesses (SMBs) with limited resources. With widely accessible generative AI tools like Duet AI, Microsoft Copilot, and Amazon’s work-in-progress AI initiatives, SMBs will be able to create professional and engaging content, and scale their business operations via automated coding and creative tools, without requiring staff with specialized skills or more tangible resources.
Google clearly recognizes this enormous opportunity in the B2B market, and was eager to showcase AI’s enterprise use cases during the keynote. For example, fast food chain Wendy’s is working with Google to train PaLM 2 to understand specific phrases and abbreviations that people say to place their orders at drive-thrus (JBC for Junior Bacon Cheeseburger, for example). Google also announced it is working with Salesforce to create an AI-powered CRM system. Interestingly, Salesforce has also been working with Microsoft to explore AI integrations for similar goals, which points to the amorphous state of the market.
One thing curiously missing from Google’s keynote is any mention of how generative AI will be integrated into YouTube’s creative toolkit. Some of the features announced during the keynote, such as the AI-powered Magic Editor in Google Photos, can easily find a use case in boosting creativity for YouTubers, yet remain unmentioned. But perhaps this is a deliberate omission on Google’s part in order to sidestep the intensifying debate around the ethics around generative AI and its potential to disrupt the creator economy. Nevertheless, as generative AI continues to take off, the key differentiating point between UGC platforms will soon be hinged around how good their integrated AI creative tools are. In the near future, the creators with the best content ideas will inevitably turn to the platform with the best AI tools to help them execute their creative vision, and YouTube can’t afford to lag behind competitors like TikTok if it wishes to stay competitive.
Human Perspectives: Tackling Copyright & Brand Safety Issues
Given the increasing regulatory scrutiny around the deployment of generative AI, exemplified by the upcoming congressional hearing on AI’s potential implications featuring OpenAI CEO Sam Altman, Google was understandably deliberate in emphasizing how it plans to curb some of the potential downsides that the mass-scale deployment of generative AI may unleash. Specifically, taking the aforementioned “bold and responsible approach,” Google promises to address some of the most pressing issues around copyright infringement and brand safety in generative AI.
For starters, Google will integrate Adobe Firefly to power Bard’s image generation feature. Adobe Firefly’s image generation tool is purposefully trained with a database of Adobe Stock images, openly licensed content from the likes of Getty Images, and public domain content, so this integration will allow users to generate images from copyrighted materials with proper attribution and licensing.
More importantly, Google is also tackling the potential abuse of generative AI tools for misinformation with new watermarking and metadata features, which embed unique identifiers and additional contextual information, such as the date and time the media was created, the tools that were used to create it, and the identity of the creator, into any piece of synthetic media content generated with Bard or other AI tools powered by PaLM 2, for easier identification. In addition, Google also says it is using adversarial testing to identify and address potential biases in its algorithms, which can lead to Bard giving out inaccurate or misleading results, as well as to identify and address potential security vulnerabilities and privacy concerns.
Lastly, to ensure that users can always get a human perspective on their questions should they need it, Google is adding a new Perspectives feature to search, which sources answers from Reddit, Stack Overflow, YouTube, personal blogs, and other community-first sites that Google’s AI deems to be the actual perspectives of actual humans. The results will be shown in a new section at the top of your search results, next to the sections of News and Images and Shopping. The goal is to help users find actual human information online and hear from other people beyond AI-synthesized results.
Through these commendable new features, Google shows that it is committed to commercializing generative AI responsibly and ethically, and therefore could be a brand-safe partner for companies that are curious to explore generative AI tools. However, some popular generative AI tools are being developed in an open-sourced nature, and Google’s announcements do not fundamentally address the misinformation and brand safety issues that generative AI faces beyond its own toolkits. Misinformation via deepfakes, in particular, will take far more human oversight and moderation than Google seemed to be committing to on Wednesday to curb in the future of synthetic media.
Hardware Priorities: Doubling Down on Foldables While Ignoring AR
Lastly, let’s take a brief look at the hardware products that Google announced during this I/O keynote. In previous years, announcements of new Pixel phones and Google Home devices would take up the majority of the stage time during the keynote. Therefore, it is quite telling that this year, with Bard and generative AI taking over the spotlight, hardware updates had become somewhat of an afterthought for Google.
That said, Google did unveil three new additions to the Pixel lineup: the Pixel Fold, the Pixel Tablet, and the budget-friendly Pixel 7A phone. The announcements for all three focused on the integration of generative AI, which will feature system-wide customization options (mainly wallpapers for now) powered by generative AI, as well as a new Tensor G2 chip that puts “AI at the center.”
The new premium Pixel 2 Fold phone, with a retail price starting at $1,799 and no wildly new features, seems to exist partly to entice developers to create Android apps with bigger resolutions that also work for the new Pixel tablet. Speaking of, the new Pixel tablet will feature a redesigned Google Home app, Google TV integration, and built-in Chromecast integration. A new Charging Speaker Dock will be available to turn the tablet into a smart home display for your kitchen counter or coffee table.
Curiously, Google made no mention of AR at all throughout the entire keynote. Considering Google’s spotty track record with immersive headset devices — remember Google Glass or Google Daydream VR headsets? — and the company’s renewed focus on generative AI, this omission is perhaps inevitable. With Microsoft and OpenAI threatening to upend the search market and challenge Google’s decade-long dominance in search ads, Google has every reason to disregard AR and focus all of its resources on maintaining parity with competitors and eliminating a potentially existential-level threat to its core business (and cash cow).
Still, with Apple heavily expected to unveil its long-anticipated AR headset device next month at the WWDC event, Google’s tunnel vision on AI, along with the resulting lack of attention towards AR, may cause it to fall behind in this rapidly developing field. In the long run, the search giant may have to play catch-up for years to come in order to keep up with its competitors in the AR market.
Want to Learn More?
This has been a special edition of our Last Weekly newsletter, bringing you the latest and hottest at the intersection of innovation and marketing. If you are keen to learn more about Google’s latest announcements and what they mean for your brand, or just want to chat about how to respond to the rise of generative AI, the Lab is here to help. You can start a conservation by reaching out to our group director, Josh Mallalieu (josh@ipglab.com).
