Without debate, 2023 could go down in history as one of the watershed moments for consumer-facing AI technology. OpenAI’s ChatGPT took the world by storm in late 2022, and every other tech giant made AI the center of their respective efforts going into 2023. Google was no different, spawning a plethora of AI-related projects for Android, its Pixel lineup, and even Google Workspace.
Although we caught a fleeting glimpse of these projects at theGoogle I/O conference in May, December is almost behind us. So, it’s time to take a retrospective look at all those ideas, because honestly, we are still waiting for widespread availability of some of that AI tech. In some cases, Google even tried to mislead the masses with some confusing, even potentially false, demoes.
If you missed it in the spring, CEO Sundar Pichai’s keynote address at I/O makes for a nice holiday season drinking game with comically numerous mentions of “AI”, and as you might imagine, the list of important launches in 2023 is similarly just as long. So, let’s dive right in.
Google Bard
Immediately after ChatGPT established itself as an acceptable way to package the endless possibilities of generative AI for the average consumer, Google launched Bard in March this year. Like ChatGPT,Bard is a conversational AI chatbotwhich uses Google’s Language Model for Dialogue Applications (LaMDA) underneath. Although it might look like a knee-jerk response to OpenAI’s efforts — and considering when Gemini launched this year, it potentially was — Bard has been in the works for years.
Its capabilities include the ability to understand and interpret natural language which humans speak, hold conversation, and exhibit contextual understanding. Bard also trumps the free version of ChatGPT with access to real-time information. Google improved Bard with LaMDA2, making it an even better remedy to your morbid curiosity about everything from arithmetic to Android and from Prometheus to plankton, and just this month, launched theGemini-powered Bardto compete with ChatGPT 4.
Search Generative Experience
Speaking of curiosity, Search is Google’s flagship offering, and it’s no surprise AI was a big part of the user experience on Search in 2023. Although Bard can answer your questions satisfactorily, its accuracy and logical reasoning stillleave us wanting for improvement. Moreover, Search has the added benefit of surfacing results from the web, which Bard only does upon request.
Fusing Bard-like generative AI capabilities with Search, Google introduced theSearch Generative Experiencethrough theSearch Labsprogram. As the name suggests, it’s purely experimental, despite its widespread availability in English. SGE precedes web links on the Search results page, offering you bite-sized nuggets of information which summarily answer your query. They can be inaccurate, but Google also adds a row of one-tap follow-up questions you can ask, just like you would with Bard. Alternatively, you can scroll past the AI-generated summary and continue visiting web links.
In its search for powerful replacements for LaMDA2 (pun intended), Google developedGemini, locking horns with OpenAI’s GPT4 language model. Gemini is primarily aimed at businesses, but it has two USPs which have us excited about the possibilities of this versatile AI model going into 2024, even if the video demo shared at launchwasn’t completely truthful.
Firstly, Gemini is multimodal, meaning it can accept prompts in any combination of text, audio, and visual media. This is like a superpower, allowing the AI to comprehend information just the way a human would use all their senses. Secondly, Gemini is available in three sizes, so to say — Nano, Pro, and Ultra. The smallest of the lot is immediately available on thePixel 8 Pro, powering features like Smart Reply in the Messages app andsummaries in the Recorder app. The mid-tier model, Gemini Pro, has been available on Bard since December, transitioning the chatbot from the older LaMDA2 model.
AI-powered features for Google Pixel devices
Gemini unlocked improved predictive response in WhatsApp for Pixel 8 Pro users, and even allowed generating summaries for recorded transcripts. Theydidn’t impress us here at AP, but Google rarely releases tech that’s perfect right out of the box. Moreover, Android 14 packs several other AI-powered improvements and features which benefit older Pixel devices compatible with the update. Here’s a quick look at two of the best bits:
Magic Compose in Messages
Android 14brings thepower of generative AI to Google Messages. Using this feature, you may respond to texts you receive rather creatively. All you need to get started is a short prompt summarizing your intended response, and a quick selection of the tone for that creative flair. AI does the rest, allowing you to deliver Shakespearean responses in a jiffy without racking your brains for those complex words.
Generative AI wallpapers
The update to Android 14 features several enhancements on the personalization front, ranging from emoji wallpapers to iPhone-like cinematic wallpapers which leverage the depth data of portrait mode shots to create a parallax effect using the device’s gyro sensors. Google also offersGenerative AI wallpapersin this version, which uses a text-to-image diffusion model. You just need to pick a subject, art style, and color palette to create a wallpaper that perfectly fits your home screen setup.
Google Workspace is where the action is
Every other aspect of Google’s AI efforts pales in comparison to the demonstrable additions to its Workspace suite of products. In 2023 alone, Google peppered apps like Docs, Sheets, and Slides with every imaginable combination of the aforementioned AI features. That’s understandable, because all these features are collectively packaged asDuet AI for Google Workspace, which costs $30 per month.
Help me Write in Gmail and Docs
Google spruced up Gmail withHelp Me Write, which elicited excitement in response to an on-stage demonstration at I/O. The feature allows you to draft thorough responses to emails without spending too much time by yourself. You just need a short prompt to get started, and the AI can refine the generated text to suit your requirements better.
The same AI is also available in Google Docs, and it operates in the same way. However, you aren’t constrained to drafting responses to the email you receive. Instead, you may use skilled prompts to create text for just about any topic you need. At one point, though, we were left feeling that Google hadtoo many text composition assistantsactive concurrently.
Help me Visualize in Slides
Since the problem of composition is much more visual when you’re crafting presentations in Slides, Google launched an AI assistant calledHelp Me Visualize, which uses text-to-image diffusion models like Dall-E and MidJourney. It effortlessly converts textual prompts into images which you’re able to drop into your presentation.
Help Me Organize in Sheets
Google Sheets is often used to manage large amounts of numerical data, and AI excels at number-crunching. To make the most of this advantage, Google’sHelp Me Organize featurein Duet AI allows you to sort data in columns and rows automatically, derive a numerical rating from customers’ text reviews, etc. Collectively, these features can be life-savers for anyone handling large amounts of data regularly.
AI in Google Meet
If the Recorder app on Pixel phones can enjoy summaries, Google Meet cannot be too far behind. Duet AI adds support for meeting summarization alongside a bunch oflighting and sound effects. The summaries can be useful when revisiting transcripts of the meeting notes, or just to save yourself some time spent attending said meetings.
That said, Duet AI can also pull data from multiple Workspace apps linked to the same Google account, simultaneously. For instance, it can take the numbers from Sheets, highlights from Docs, and meetings notes from Meet, and throw it all into one impressive Slides presentation poised to blow your manager away. Duet AI is also available for Google Cloud customers.
Noteworthy mentions
While most of the consumer-facing AI products Google launched fall into one of the above segments, there have been a few wildcard creations this year. They have a promising future ahead of them if Google chooses to continue developing them.
Immersive View on Google Maps
What started out as a rival for the 3D view in Apple Maps is now calledImmersive View in Google Maps. The feature uses AI to create immersive 3D representations of places you’ve only dreamt of visiting, using just the flat images of the location. Understandably, availability is limited to a handful of cities, but it’s a cool use case for AI where realistic representation of the surroundings can help you plan and navigate better. Until availability expands, though, it’s just another AI pipe dream.
Imagen Editor
AI already plays a big role in the image editing process for Google Photos users. However, Imagen AI’s editing suite showcased this year practicallyredefines effortless editing. Think the convenience of Bard combined with the accuracy of Adobe Photoshop’s Content-Aware Fill feature. Imagen editor specializes in localized edits which affect only a defined region of the image. Once the mask is defined, textual prompts suffice and AI ensures the edit is done cleanly, saving you the trouble of going in with a fine-toothed comb later. However, it isn’t immediately available to the public.
NotebookLM
Alongside the recent Gemini debut, Google announcedNotebookLM, an AI tool previously called Project Tailwind, powered by the Gemini Pro model. Its singular objective is to simplify note-taking and content formatting in said notes. AI can assist users in organizing haphazard notes, thanks to contextual understanding and information processing capabilities. In theory, Google only designed this for students, but they aren’t the only people planning their work and taking notes. NotebookLM can benefit millions if deployed at scale.
Sliding into 2024 with hopes and concerns
While everyone is excited to see the possibilities of AI unravel, there are several boundaries which lack definition. It’s no secret that AI uses vast amounts of data to improve, and suchdata collection raises privacy concerns. Moreover, there’s a massive risk of plagiarism and unoriginal content withAI-generated work closely resemblingthe human work it is trained on — Google’s gone as far as topromise to protect its usersfrom lawsuits directly. People can also pass off AI-generated content as their own, raising even more ethical concerns. Whilewatermarkingof images and text created by AI could help matters, we hope companies like Google get behind such initiatives.
