New Google AI Products Taking On Amazon And Microsoft

At the Google I/O conference, which kicked off on May 10 at its headquarters in Mountain View, California, Google’s chief executive officer Sundar Pichai did his best to assure developers, partners, regulators and the stock market that the tech giant intends to be “bold and responsible” in the use of generative AI.

A bad demo in February cost the company $100 billion in market value, but Google ended on a high note on Wednesday after presenting a well thought out response to Microsoft’s early gains in the AI wars with ChatGPT-maker OpenAI.

“Seven years into our journey as an AI-first company, we’re at an exciting inflection point,” Pichai said as he highlighted features rolling out across the product roadmap. With 15 products each serving more than half a billion people a year and six serving over two billion each, including Search, Gmail, Android, Chrome, YouTube and Drive, Google is seeking to make AI helpful for everyone, he said.

Here are the new AI products and features to watch for:

  • In what appears to be a direct challenge to Amazon, Search is getting an AI-powered Snapshot section with a detailed summary of options, pricing and reviews to respond to shopping prompts like “What’s a good bike for a five mile commute with hills?”
  • Gmail is getting a Help Me Write tool that can compose a range of responses to emails that include styles that are formal and elaborate, short and sweet, and “I’m Feeling Lucky” for those unsure of how to respond. This builds on the Smart Reply tool released in 2017 which provides quick responses to emails, and the Smart Compose tool that followed which offers writing suggestions as users type. The Help Me Write tool has been available in both Gmail and Docs to “trusted testers” since March.
  • Maps is getting Immersive View for Routes, expanding on a feature introduced at last year’s Google I/O to provide scenic views of biking, walking and driving routes that users can zoom in to see points of interest, weather conditions, air quality and traffic. Coming to 15 cities by the end of the year including New York, San Francisco, Tokyo and London.
  • Photos is getting Magic Editor which can fill in the blanks of things cut-off in a photo and builds on Magic Eraser which can delete items. One of Google’s first native AI products, Photos was introduced in 2015 and uses machine learning to search photos for people and things, like sunsets. Magic Editor is expected later this year.

  • Catching up to ChatGPT, Bard can now code with citations, debug and explain snippets in 20 languages including ASP.NET, C, C++, C#, Dart, Go, Google Sheets, Groovy, HTML/CSS, Java, Java Script, Kotlin, Matlab, PHP, Python, R, Ruby, Rust, SQL and Swift powered by Google’s newest large language model, PaLM 2. And soon the chatbot will be able to export responses into Maps, Docs and Sheets as well as other Google products. It’s not clear whether Bard will be providing links along with its responses. Oddly the keynote demos didn’t show any links and in a conversation I had with Bard in April, the chatbot told me that if I wanted links, I needed to Google them.
  • In the coming months, Bard will be partnering with a number of companies including Spotify, Instacart, Walmart, Zillow, Redfin, Khan Academy, ZipRecruiter, Indeed, TripAdvisor, OpenTable, Kayak, Uber Eats, Data Commons, Wolfram, FiscalNote, Replit and Adobe for its AI art generator, Firefly.
  • In addition to Bard, PaLM 2 is the foundation model for 24 other products including Gmail, Docs, Sheets and YouTube, and was trained on more than 100 languages. It is being released in four sizes whimsically named Gecko, Otter, Bison and Unicorn, with Gecko the lightest for mobile. The model is customizable for specialized domains like Med-PaLM 2 for medical applications and Sec-PaLM 2 for security. Pichai also teased Gemini, a large language model in training to be multimodal for memory and planning being developed by the newly combined Google Brain and DeepMind teams, now referred to as Google DeepMind.
  • Duet AI for Workspace is coming later this year with an always-on AI collaborator tool that proactively offers prompts, speaker notes for Slides, and Sidekick for text-to-image generation that can create both illustrations and photorealistic images.
  • Duet AI for Cloud is releasing three Vertex AI models for enterprise clients: Imagen for image generation, Codey for coding, and Chirp for speech-to-text recognition. Companies currently using the models include Uber, Wendy’s, Deutsche Bank, Replit, Orange and Canva. Canva has a Magic Video feature in the pipeline to pull together stories from uploaded media. And according to CNBC, Wendy’s will be piloting an AI-powered drive thru in Columbus Ohio in June and featuring a chatbot that can understand Gen Z orders as described by Wendy’s CEO Todd Penegor in a video testimonial during the keynote. Duet AI also offers enterprise search similar to services offered by Sequoia-backed unicorn startup Glean founded by former Google engineer Arvind Jain.
  • Most exciting, in the works for Drive is an AI notebook codenamed Project Tailwind, which will enable source materials to be dropped into a folder and synthesized into study guides, earnings reports, news articles and perhaps even legal documents. To preview, sign up at g.co/labs.
  • Of course, with all of this AI-generated media, users are going to need help understanding what’s real and what’s fake so the About This Image tool in Search will start to show provenance as to when the image might have first appeared and where else it’s been seen online. And as Google rolls out its image generation features, media will be tagged with metadata that discloses when an image has been AI-generated. Additionally, Google is working on digitally watermarking the images it creates.

Updated with additional PaLM 2 details in paragraph 11.

Read more here: Source link