AI developement options
1. AI developement options
SPEAKER: In the previous module, you explored cutting-edge technologies about generative AI. But what if you want to create predictive AI for traditional AI tasks like forecasting and classification? How do you build an ML model, and what options do you have? Find the answers in this module. You begin by comparing AI development options on Google Cloud, from no-code to low-code, and finally a do-it-yourself approach. These options apply to both Gen AI and predictive AI. You are then introduced to Vertex AI, Google Cloud's unified AI development platform, and your playground to build an ML model from end to end. After that, we take a deep dive into how each of three options works-- AutoML, pre-trained APIs, and custom training. Never miss a practice. You conclude with hands-on experience using the natural language API to identify subjects and analyze sentiment in text. Let's start with Google Cloud's AI development options. What choices do you have for both generative AI and predictive AI projects? And how do you make the right decision? Let's find out. Imagine transforming your organization's business model and operations with AI. Perhaps you're a business user without a technical background, but you're eager to prototype business ideas using AI. What are your options? Or maybe you're a data scientist with training data looking to build a custom ML model without spending hours tuning parameters from scratch. What choices do you have? Even if you are an ML engineer who enjoys a do-it-yourself approach to building ML pipelines, what tools can you utilize? Google Cloud can help you achieve your goals by meeting you where you are, offering no-code, out-of-box solutions like Gemini Enterprise and Conversational Agents introduced in the previous Gen AI module-- no- to low-code solutions like AutoML, which helps you build your own ML models through point and click; low-code solutions like pre-configured APIs, which use pre-trained ML models, eliminating the need to build your own if you lack training data or in-house ML expertise, code-based approaches, ranging from BigQuery, ML, and Agent Development Kit, ADK, both introduced earlier, to completely custom training. Beyond technical expertise, how do you choose the right tool to build an ML model? This brief guide comparing four options-- pre-trained APIs, BigQuery ML, AutoML, and custom training may offer some insight. BigQuery ML only supports tabular data and semi-structured data like JSON files. AutoML supports tabular and image data. Whereas the other two support tabular data, images, text, and video. Pre-trained APIs also process audio. In terms of training data size, pre-trained APIs do not require any training data, whereas BigQuery ML and custom training require a large amount of data. Pre-trained APIs and AutoML are user-friendly, with low requirements for machine learning and coding expertise. Whereas custom training has the highest requirement, and BigQuery ML requires you to understand SQL. At the moment, you can't tune the hyperparameters with pre-trained APIs or AutoML. However, you can experiment with hyperparameters by using BigQuery ML and custom training. Pre-trained APIs require no time to train a model, because they directly use pre-trained models from Google. The time to train the model for the other three options depends on the specific project. Normally, custom training takes the longest time because it builds the ML model from the beginning, unlike AutoML and BigQuery ML. The best option depends on your business needs and ML expertise. Budget is also an important consideration. Visit Google Cloud's website for detailed pricing information. If you have little ML experience and no intention to train your own ML models, using pre-trained APIs might be the best choice. Pre-trained APIs address common perceptual tasks such as vision, video, and natural language. They are ready to use without any model development effort. If your data engineers, scientists, or analysts are familiar with SQL and already have data in BigQuery, BigQuery ML lets you use SQL queries to build pre-defined ML models. If you wish to build custom models with your own training data while you spend minimal time coding, then AutoML on Vertex AI is your choice. AutoML allows you to focus on business problems instead of the underlying model architecture and provisioning. If your ML engineers and data scientists want full control of the ML workflow, Vertex AI custom training lets you train and serve custom models with code on Vertex AI Workbench or Colab Enterprise. Before delving into these options, the next lesson will introduce Vertex AI, Google Cloud's AI development platform.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.