Choosing an AI solution
Buy what is common, and build what sets you apart
From a pre-trained API in hours to a custom model in months: speed falls as differentiation rises.
There are four ways to get an A I capability, and they trade speed for uniqueness. A pre-trained A P I is a finished skill. You bring only the request, it works in hours, and everyone else can buy the same thing. A foundation model such as Gemini needs a good prompt, and days. A tuned or Auto M L model is trained on your own examples. It takes weeks, and it knows your business. A custom model is built from the start with data, code and specialists. It takes months, and it is yours alone.
Choosing an AI solution · Going deeper
Five things to weigh
Speed, effort, differentiation, expertise, and choice and flexibility.
Pre-trained APIs
A pre-trained API is a finished skill that you call
The Vision API sees, Speech-to-Text hears, Cloud Translation translates, and Gemini does many things.
Start with what needs no training at all. The Vision A P I takes an image, such as a photo of a receipt, and returns what is in it: the shop, the date and the total. The Speech to Text A P I takes a recording and returns a transcript. The Cloud Translation A P I takes text in one language and returns it in another. Gemini is different in kind. It is a foundation model: give it a question with a video attached, and it writes an answer.
Pre-trained APIs · Going deeper
APIs and foundation models
Single-purpose APIs do one job. A foundation model does many from a prompt.
Check yourself
A contact centre wants what the drawing shows. Which product belongs at the "?"
The Agent Platform
One platform, from choosing a model to running an agent
Gemini Enterprise Agent Platform: choose a model, use Agent Studio, build an agent, then run and govern it.
Calling a model is the easy part. Building something a business can rely on takes a platform, and Google calls its one Gemini Enterprise Agent Platform. First you choose a model: Gemini, or one of many others. In Agent Studio you test prompts, and tune the model with your own data. You build an agent by connecting it to tools and to company data. Then you run it: deploy, scale, watch and secure it. The platform also holds Auto M L, for custom models without code, and the A P I your applications call.
The Agent Platform · Going deeper
What gets built on it
Service agents, search over company knowledge, content, code help and predictions.
Custom models
Three ways to teach a model your own data
BigQuery ML uses SQL, AutoML uses no code, and Agent Studio tunes a foundation model.
When a ready-made model is not enough, you train on your own data. The guide names three ways. BigQuery M L, for people who know S Q L. One statement creates a model from a table, and the data never leaves the warehouse. Auto M L, with no code. You supply labelled examples, and the platform trains a model for you. Agent Studio, starting from Gemini. You give a foundation model your own examples, and get one tuned to your business.
Custom models · Going deeper
The three compared
"Standard SQL" means BigQuery ML. "Without code" means AutoML.
Check yourself
A retailer describes its situation as shown. Which product fits best?
AI Hypercomputer
Chips, software and ways to pay, designed as one
The AI Hypercomputer: GPUs and TPUs, open software, and flexible consumption to control cost.
Under every model is the computing that trains and runs it. Google calls its system the A I Hypercomputer, and it has three layers. Hardware: G P Us, and T P Us, the chips Google designs only for A I. Software: open frameworks and open standards, tuned to those chips. And consumption: you can pay on demand, reserve ahead for less, or use spare capacity. Because the three are designed together, models train and answer faster, at a cost you can control.
AI Hypercomputer · Going deeper · 1 of 2
GPUs, TPUs and ways to pay
A TPU is Google's own AI chip. Committed use and spot capacity lower the bill.
AI Hypercomputer · Going deeper · 2 of 2
Clue words for AI offerings
Decide how much to build, then match the phrase to the product.
Check yourself
A company describes its need as shown. What should it do?
Lesson complete
You finished “Google Cloud AI offerings”
Answer the questions along the way to see your score here.