Google is attempting to reclaim its crown as the leader in artificial intelligence with PaLM 2, a “next-generation language model” that the company says outperforms other leading systems on some tasks.
Unveiling cutting-edge AI at its annual I/O conference, along with a foldable Pixel phone and a new tablet, Google said it would be incorporated into 25 new products and features as the company rushes to catch up with competitors. after years of producing AI research but few products.
Like other “big language models” like OpenAI’s GPT, PaLM 2 is a general-purpose AI model, which can be used to power ChatGPT-style chatbots, but also to translate between languages, write computer code, or even Analyze and respond to images. Combining those capabilities, a user could ask a question in English about a restaurant in Bulgaria, and the system could search the web for Bulgarian answers, find an answer, translate the answer into English, add a picture of the location, and then move on with a code snippet to create a database entry for the place.
“The neural network revolution that we are experiencing now started about 10 years ago,” said Slav Petrov, co-leader of the PaLM 2 project, “and it started in part at Google.” Advances in AI, including the “transformer”, the T in GPT [Generative Pre-Trained Transformer]came from the company’s investigation, Petrov said.
“We are very excited to make these models widely available externally, because we want to see what people can do with them,” he added. “We believe they will open up many opportunities to do things that were once considered magical and truly out of reach, but are now achievable thanks to the incredible progress in machine learning that we have seen in recent years. ”
The most obvious way to interact with PaLM 2 will be on Google’s own chatbot, Bard, which is opening up to the general public for the first time and rolling out globally. Taking full advantage of PaLM 2’s multilingual capabilities, Bard is also available in Japanese and Korean, as well as English, and the company intends to support 40 languages in time.
Chatbot users can also submit photos of Bard for the first time, and the company gives an example of submitting a picture of a kitchen shelf and asking for a recipe with the ingredients. In a reversal of the standard, which replicates a feature promised by OpenAI in conjunction with the release of its latest and most powerful AI model, GPT-4, but is not yet generally available, leaving Google to the head in the so-called “multimodal capabilities”.
In a new feature the company calls “Duet AI,” users of Google’s “Workspace” apps (Gmail, Docs, Slides, and Sheets) will also be able to use PaLM 2 AI as a co-author of text, spreadsheets, and slides. . An image generator built into Google Slides, for example, will let you task an AI to visualize your ideas, while a “help me write” button in Google Docs can generate entire swaths of text automatically. In one example, the “regional sales representative job opening” ad quickly expanded into a full job description, replete with clear spaces to enter specific details, such as company name and location.
But a disclaimer that the tool “is a creative writing aid and does not claim to be objective” underpins the dilemma for Google: Rushing technology to beat the competition also carries risks of misbehaving AI software. The launch of the AI system was made at the last minute, with detailed updates sent to reporters just hours before the company’s chief executive, Sundar Pichai, took the stage at the conference.
In its preliminary research, the company warned that PaLM 2-based systems “continue to produce toxic language damage,” with some languages producing “toxic” responses to queries about black people in nearly one-fifth of all tests. , part of the reason why Bard chatbot is only available in three languages at launch.