On May 10, 2023, Google launched its latest generative AI language model, PaLM 2. This LLM is the successor of its previous language model, PaLM. Currently, PaLM 2 is a multilingual model with advanced reasoning and coding capabilities.
According to Google, Bard is powered by PaLM 2. So, you can access PaLM 2 functionalities in Bard.
But what if you want to interact directly with PaLM 2’s interface? Well, it’s possible via Vertex AI. Continue reading this post to learn more about how to use PaLM 2 AI model using your Google Cloud account.
Step 1: Set Up Google Cloud Account
The first step to using Google PaLM 2 and its functionalities is to create a Google Cloud account. You can create a free account and get $300 worth of free credit.
Follow the steps given below to set up your Google Cloud account:
- Open your web browser and visit the Vertex AI (https://cloud.google.com/vertex-ai) website.
- Click on the Try Vertex AI for free button.
- Sign in with your Google account credentials.
- Provide your business details (name and size of the organization).
- Accept the Terms of Service and click on Continue.
- Enter your payment details. This step is mandatory since Google adds free credits to your account and charges you after you exhaust the free limit.
- After completing the payment verification step, you can use your Google Cloud account.
Step 2: Access the Vertex AI Platform
You can access Vertex AI within your Google Cloud account. The following steps will help you get started:
- Click on the search bar and type Vertex AI.
- Choose Vertex AI under the products and pages list.
- Select the Language option under the Generative AI Studio option.
- You will see three labels on your screen. Choose the Create Chat Prompt in the middle label.
- Type a prompt in the space provided. It will return an error asking you to enable the API.
- Hit the Enable option to start using the Vertex AI. Close the popup after enabling Vertex AI.
Step 3: Use the Google PaLM 2 AI Model
You are now interacting with the Google PaLM 2 model. Ask the model any question using its text bar. You can ask it to write code, translate text, or summarize a paragraph.
You will see various setting options on the left side of your screen. Adjust them to get responses of different levels and styles. You can change the temperature, token limits, and Top-P and Top-K values.
PaLM 2 is available in four different models based on size and performance. Gecko is the smallest while Bison is the largest model.
You can switch between different models using the Model option in the left-side menu. Currently, PaLM 2 is using the Bison model by default.
PaLM 2 (Bison) AI Model: First Impressions
Bison is a small and fast subversion of PaLM 2. The model is trained with data upto February 2023 and will undergo improvements in the future.
The current version is much smaller than Open AI’s latest LLM, GPT-4. Besides, it is excellent at coding and beats other language models in various aspects.
Bison hallucinate much lesser when asked fictional and reasoning-based questions. It understands riddles and idioms accurately. It also interprets and solves complex queries much more efficiently.
PaLM 2 is trained with a large set of data in various languages, making it excellent at translating text and understanding different languages.
