“Google Is Beta Testing Its AI Future: After Mistakes and Challenges, the Company Is Moving a Little Slower With AI Language Models”, 2022-05-11 ():
It’s clear that the future of Google is tied to AI language models. At this year’s I/O conference, the company announced a raft of updates that rely on this technology, from new “multisearch” features that let you pair image searches with text queries to improvements for Google Assistant and support for 24 new languages in Google Translate.
…Step into the AI Test Kitchen: Over a video call ahead of I/O, Josh Woodward, senior director of product management at Google, is asking Google’s latest language model to imagine a marshmallow volcano.
“You’re at a marshmallow volcano!” says the AI. “It’s erupting marshmallows. You hear a giant rumble and feel the ground shake. The marshmallows are flying everywhere.”
Woodward is happy with this answer and prods the system again. “What does it smell like?” he asks. “It smells likes marshmallows, obviously”, the AI replies. “You can smell it all around you.” Woodward laughs: “Okay, so that one was very terse.” But at least it made sense.
AI Test Kitchen will let select users test Google’s latest language model: Woodward is showing me AI Test Kitchen, an Android app that will give select users limited access to Google’s latest and greatest AI language model, LaMDA 2. The model itself is an update to the original LaMDA announced at last year’s I/O and has the same basic functionality: you talk to it, and it talks back. But Test Kitchen wraps the system in a new, accessible interface, which encourages users to give feedback about its performance.
As Woodward explains, the idea is to create an experimental space for Google’s latest AI models. “These language models are very exciting, but they’re also very incomplete”, he says. “And we want to come up with a way to gradually get something in the hands of people to both see hopefully how it’s useful but also give feedback and point out areas where it comes up short.”
…The app offers 3 demos that showcase what LaMDA 2 can do. The first is a simple brainstorm tool that asks the app to help you imagine if you were in various scenarios. During the keynote demo, Google entered “I’m at the deepest part of the ocean” as a response to the app’s prompt of “Imagine if.” The app then spit out a short paragraph describing the user in a submarine the Marianas Trench, with descriptive language.
Secondly, as a demonstration of the model being able to stay on topic, the app can have a conversation with you about something and understand context. During the demo, the app started by asking “Have you ever wondered why dogs like to play fetch so much?” In its responses to simple follow-ups like “Why is that”, the system replied with more information about dogs and their senses of smell.
Finally, AI Test Kitchen shows how LaMDA 2 can “break down a complex goal or topic.” This section is called List It, and users can ask things like “I want to learn ukulele” or “I want to grow a garden.” LaMDA will generate lists of subtasks to help you get started, and according to Google, may even offer ideas you might not have thought of. In addition to giving you the names of vegetables you can grow, for example, AI Test Kitchen might also give you a set of steps to take or weather conditions to consider. During the demo, the app offered a tip for users with limited space, sharing the types of plants that might thrive in smaller gardens.
…In addition to LaMDA 2, Google also released AI Test Kitchen, a center for AI demonstrations driven by models like LaMDA 2. As an app, users can interact with the model in limited ways, such as exploring a specific topic (like a dog), and drilling down to subtopics (a dog’s smell). Google says it will continue to add “other emerging AI fields” to AI testing kitchen, including the field of natural language processing and others.
AI Test Kitchen will launch in the US in the coming months, but not publicly. Google has not yet fully decided how it will provide access, and the company is weighing reaching out to select academics, researchers and policymakers.
[Imagine looking at all the stuff being done with OA API or FB etc, and seeing that this is how Google is rolling out LaMDA (never mind Chinchilla or PaLM or…). No wonder everyone is leaving.]