Google Cloud Unveils New AI Tools: PaLM API, MakerSuite, And Generative AI App Builder, To Revolutionize App Development
How Google Cloud is empowering developers with cutting-edge AI for next-gen applications.
Google Cloud Unveils New AI Tools
Google Cloud has announced a range of new AI-powered features for its productivity tools, including APIs and tools for developers to create AI-powered applications. The company is also launching MakerSuite, a browser-based tool that simplifies the development of AI applications on top of Google’s foundation models. Additionally, Google is integrating support for generative AI into Vertex AI, its platform for building and deploying ML models, and launching the Generative AI App Builder to aid developers in creating bots, chat interfaces, digital assistants, and custom search engines.
Google’s new AI features will initially be available to a limited number of developers and only accessible through Google’s North America data centers. Two big-name customers already using these tools are Deutsche Bank and Toyota. The core of these new developments is the PaLM API, which provides developers with access to Google’s large language models, enabling them to build with generative AI.
As per Google Cloud CEO Thomas Kurian – the new API is an “extremely approachable way for developers to start building with generative AI.” Developers can access foundation models, tune and augment them with their own data, and utilize synthetic data to build applications responsibly and safely. Google plans to offer an “efficient model available in terms of size and capabilities” and add other models and sizes soon.
For developers who prefer not to use the API, Google is launching the low-code MakerSuite service. Initially available to Trusted Testers, MakerSuite will offer two models: PaLM chat-bison-001 for building chat-style, multi-turn applications, and PaLM text-bison-001 for single-turn input/output scenarios.
Google also discussed Vertex AI and the Generative AI App Builder in detail. Vertex AI provides an end-to-end service for developers to build AI models and applications, and now includes access to foundation models for generating text and images. Audio and video generation will be added in the future. The Generative AI App Builder is a new service enabling developers to create AI-powered chat interfaces and digital assistants based on their own data.
To accomplish this, they combined foundation models with enterprise search capabilities and conversational AI for building multi and single-turn conversations. Kurian explained that this tool can be used to retrieve information and transact, while also allowing users to control the generative flow.
Google emphasizes that a company’s training data will always remain private and not be used to train the broader model. The focus is on business users who want to augment the model with their own data and/or fine-tune it for their use cases.
While Google’s well-known LaMDA model has not been publicly released, the company opted to use the PaLM model as the foundation for these services. Launched a year ago, PaLM has 540 billion parameters, making it significantly larger than OpenAI’s GPT3 with 175 billion parameters. With ChatGPT-4 already released, it is interesting to see how PaLM and Google’s AI initiatives will compete and evolve alongside the GPT series. The implications of this choice, particularly in light of GPT 3.5 on the market and GPT4’s successful launch, remain to be seen. Google’s decision to use PaLM may be due to its strong performance across coding tasks and natural language tasks, even with only 5% of code in the pre-training dataset.