During its annual Google I/O conference, internet giant Google unveiled a slew of new AI-powered capabilities, with updated AI technologies as they appear across its major platforms. The annual Google I/O conference was held in California on May 10, with CEO Sundar Pichai delivering a keynote address on the most significant changes to the company’s AI stack, among other announcements.
Google introduced the Pathways Language Model (PaLM) in August. Developers have since leveraged language learning techniques to release generative AI-related applications such as the ever-popular chatbots.
Google’s model, “PaLM 2,” is updated with improvements to reasoning, coding, and multilingual capabilities as it is trained on increasingly difficult and varied topics.
The PaLM 2 will be available in multiple sizes, with one iteration of the device capable of being used on mobile. Google stated that the new PaLM forms the foundation for more than 25 applications and demonstrated two specific models. Sec PaLM 2 is used in cybersecurity while Med-PaLM 2 is used in medical applications. As a competitor, Google’s core Search product gets AI support. Microsoft’s Bing beats Google by including OpenAI’s ChatGPT in its search engine.
Search Generative Experience (SGE), as it’s known colloquially, will have a limited trial rollout to US users opting in for the test before Google considers larger adoption. According to product demonstrations, the tool appears to collect information from web pages and provide it in ChatGPT-like answers above the actual search results in Google Search.
When consumers use the “shop” option in Search, it also provides information about the product they are looking for. When a user searches for e-bikes, for example, the model makes suggestions about what features to look for in a bike, according to Google’s presentation.
Bard, Google’s answer to ChatGPT, was among the items given the PaLM 2 makeover, which included new functionality and a wider rollout. The conversational AI model was first introduced roughly two months ago, but only in the United States and United Kingdom; now available in over 180 countries, with more to come.
Improvements to Bard’s coding capabilities and repertory are part of the upgrade. Google has improved its citations as well, with the bot noting where it obtained certain codes. Adobe’s AI generative image tools will soon be integrated into Bard, allowing it to generate images with prompts, similar to other popular applications.
Many additional Google products are receiving PaLM 2 support, and Pichai is leading a series of demonstrations highlighting new AI-powered features in Google Maps, Docs, Mail, and Photos. One example showcases Gmail’s AI-powered version of “Smart Compose” feature, which can compose email responses on request. It also seems to pull data from email threads to segment responses and can be modified to make the wording more formal, detailed, or concise.
A similar app, “Magic Compose”, is coming to Google’s Android phones with AI-generated responses that supposedly help make messages express “desired vibes” like “chill” or “Shakespeare”. Despite the fact that PaLM 2 was just released, Google is working on a supposedly more advanced big language model called “Gemini” to replace it or, at the very least, provide other options.
Although Gemini is still in the early stages of development, Pichai stated that Google is “already seeing impressive multimodal capabilities not seen in previous models”. He noted that once “well-tuned and rigorously tested for safety”, the Gemini, like the PaLM 2, will be available in a variety of sizes and capabilities.