Google has announced a number of features to stay ahead in the AI race, especially competition from Microsoft and OpenAI. The search engine will look a lot different in the coming days as it will integrate "meaningful" AI. Here’s a look at some of the interesting developments that came out of Google I/O 2023. Google has spent billions of dollars over the years and it appears determined not to lose the race to any rival company.
Search Generative Experience
The search engine will never be the same. To put it mildly, it will no longer be boring. Google is introducing something called Search Generative Experience (SGE), which can be accessed via a Google Labs waitlist (at the moment, sign-ups are allowed only in the US).
You will still be typing in your queries but the results are displayed differently. It’s using generative AI that can answer almost like a human. SGE is not Bard, which is another AI product from Google. It’s not Google Assistant, which is the bot that we interact with our voice.
Once a query is given, users will see suggested next steps as if we are conducting a research with an AI-powered snapshot of information. There will be links to go deep into a subject. Tapping on a suggested next step takes the search into a conversational mode where you can ask more questions on the topic that’s being explored, while context is carried forward from the previous query.
At the moment, completing online research on a topic requires multiple searches and visiting individual websites. We end up with a bunch of tabs open on Chrome. Google’s AI can summerise facts and ideas from different parts of the Internet.
Not only that, corroborations are included, that is, links to several sources are shown on the left side, clicking on which will expand into a view of the source websites. Plus, you don’t have to break up the initial search query into smaller queries to get a valid answer.
Where Google trumps Microsoft is its integration with shopping websites and the wide range of apps that it has had on offer for a long time. For example, if you are looking for a new bike, you will get answers that result from the integration with Google Shopping. So you will be shown factors to consider while buying a bike and your follow-up questions will refine the search further. The new generative AI shopping experience is built on Google’s Shopping Graph, which has more than 35 billion product listings, Google has said.
No waitlist for Bard
Google’s AI chat bot is free to speak. The company has said that the chat bot is now available to everyone and there is an addition of new features, such as support for Japanese and Korean languages and an easier way to export generated text to Google Docs and Gmail.
The question is when should we use Google Search, which now has more AI power, and when do we use Bard? Bard is a chat bot that can hold conversations and it can be used, for example, to generate software code or write a caption for a photo. As for traditional Google search, it will help with finding and seeking information. What you will notice is a difference in the answers. If the new Google detects that generative AI can be used to help with a query, the top of the results page will show the AI-generated response. The traditional links to the web will remain below.
Google plans on adding more functionality to Search, including AI image generation that uses Adobe’s AI image generator, called Firefly and integrations with third-party services like OpenTable and Instacart.
PaLM 2
It’s one of the most important ingredients in the AI map. PaLM 2 is the newest large language model (LLM) that will power Google’s updated Bard chat tool, the competitor to OpenAI’s ChatGPT. At the moment, PaLM 2 is now available to developers through Google’s PaLM API, Firebase and on Colab.
“PaLM 2 is a state-of-the-art language model. It’s good at math, coding, reasoning, multilingual translation and natural language generation,” said Zoubin Ghahramani, a vice-president at Google’s DeepMind division that oversees the company’s artificial intelligence work.
The model is being used with 25-odd Google products, including Google’s Bard chat bot, Gmail, Google Docs, Google Sheets and YouTube. The only technical details Google has offered are that PaLM 2 has been built on top of Google’s latest JAX and TPU v4 infrastructure.
The model has been trained on 20 programming languages, including JavaScript and Python, Prolog, Verilog and Fortran. It forms an integral element of Codey, Google’s specialised model for coding and debugging.
At the same time, Google is training a successor called Gemini, which is going to be multimodal, “highly efficient at tool and API integrations and built to enable future innovations, like memory and planning”.