When you log in to ChatAible you’ll notice it looks a little different.
Now you see something like this:
Instead of starting with Datasets you can launch other Aible capabilities via the New Chat page.
Use Cases
ChatAible lets you interact with different use cases and we’ve released 3 examples to begin with with more to follow - and in the future you’ll be able to create your own.
- the Analytics use case interacts with your datasets - using Aible’s advanced information model to inform and guide the generative AI to provide high level analytical answers
- the ChatAible Help use case lets you summarise and query the content of this community site. This is an example of document summarisation and we will be adding the capability for you to interact with your own documents soon.
- lastly, the Anything use case is just what it says: you can ask Aible anything with no prompt engineering applied to your request.
Tones
We’ve separated out the tones from the use case so you can control the style of the LLM responses independently of the content.
In the future you’ll be able to add your own tones:
Choose your LLM
We’ve been exerimenting with different language models for a while now and are pleased to be making Google PaLM 2 available to all ChatAible users:
You can see from the icons some relative comparisons of the strengths of each model:
- OpenAI GPT-4 is a larger model which generally gives better quality results but it takes a lot longer and at a higher cost per token
- Google PaLM 2 on the other hand yields much quicker responses but often of a lower quality.
We will be adding more models over the coming weeks and months as new capabilities become available.
Data Discovery
Using the Analytics use case you can use the language models to find data relevant to your query.
For example, here I’ve asked ChatAible What data do we have related to sales?…
…notice how it has returned a list of datasets that have content related to my question.
I can click on any of the datasets to ask ChatAible to prepare an executive summary (using the Tone and Language Model I’ve selected).