The Googlewho held I/O Developers Conference 2022 Wednesday afternoon folded up On the development of artificial intelligence (AI) and machine learning (ML). It focuses not only on research, but also on product development.
One area of Google’s work is to make its products, especially communications products, “more accurate and natural.” This includes the development and implementation of new language processing models.
Take a look at what the company announced:
AI Test Kitchen
After launching LaMDA (Language Model for Dialog Apps) last year, which allowed the Google Assistant to conduct more natural conversations, Google announced LaMDA 2 and AI Test Kitchen, an app that gives users access to this model.
AI Test Kitchen will allow users to explore these AI features and give them an idea of what LaMDA 2 can do.
Google launched an AI Test Kitchen app with three demos: first, titled “Imagine It,” allows users to suggest a conversational idea, and then Google’s language processing model returns “imaginative and relevant descriptions” about the idea. The second, called “Speak It,” ensures the language model stays on topic, which can be challenging. The third form, called List It Out, will provide a list of potential to-dos, things to watch out for, or pro tips for a particular task.
Paths Language Model (PaLM)
PaLM is a new model for natural language processing and artificial intelligence. According to Google, this is its largest model to date and has been trained on 540 billion parameters.
Currently, the model can answer math problems or explain a joke, thanks to what Google describes as a thought chain, allowing it to describe multistep problems as a series of intermediate steps.
An example shown with PaLM is an AI model that answers questions in Bengali and English. For example, Sundar Pichai, CEO of Google and Alphabet, asked the model about popular New York pizza toppings and the answer came up in Bengali, though PaLM never saw parallel phrases in the language.
Google hopes to expand these capabilities and technologies to include more languages and other complex tasks.
Multiple search in Lens
Google has also announced new improvements to its Lens Multisearch tool, which will allow users to search only with an image and a few words.
“In the Google app, you can search by image and text at the same time, and you can also point to something and ask a friend about it,” the company said.
Users will also be able to use a photo or screenshot and add “Near Me” to see options for local restaurants or retailers carrying clothing, household goods, and food, among other things.
Through a hack called Scene Scan, users will be able to use Multisearch to pan the camera and instantly get information about multiple objects in a larger scene.
Immersive Google Maps
Google has announced a more comprehensive way to use its Maps app. Using computer vision and artificial intelligence, the company has combined billions of Street View and aerial imagery to create a rich digital model of the world. With the new comprehensive view, users can experience what a neighborhood, landmark, restaurant or popular spot looks like.
Google has also added 24 new languages for translation, including Assamese, Bhojpuri, Konkani, Sanskrit, and Meso. These languages are added with Zero-Shot Machine Translation, the machine learning model only sees monolingual text, which means it learns to translate into another language without seeing any examples.
However, the company indicated that the technology is not perfect and that it will continue to improve these models.
“Problem solver. Proud twitter specialist. Travel aficionado. Introvert. Coffee trailblazer. Professional zombie ninja. Extreme gamer.”