4 ms·
You've inspired me... here is what GPT-3 has to say about Infinidash (prompt (including first newline) in italics, first take): Introducing AWS Infinidash Inf
by colanderman 5y ago
You've inspired me... here is what GPT-3 has to say about Infinidash (prompt (including first newline) in italics, first take):
Introducing AWS Infinidash
InfiniDASH is a container-based, serverless, infinitely scalable, high-performance, low latency event stream processing platform built on top of AWS Lambda, Kinesis, and Amazon DynamoDB.
Applications built with InfiniDASH can handle millions of events per second with sub-millisecond latency.
InfiniDASH is currently in developer preview, so you can try it out for free.
To learn more about InfiniDASH, visit the product page.
To see a live demo of InfiniDASH at re:Invent, register for a hands-on lab.
— Jeff;
- bpodgursky 5y agoSo... now I'm wondering whether this actually should be a thing. That sounds pretty slick.
- jcims 5y agoThat's hilarious. I think Jeff has an account here. Just tried with GPT-J posted here yesterday(?) Prompt in italics as well. Seems like it wants to tweak the name: Introducing AWS Infinidashboard What is AWS Infinidashboard? AWS Infinidashboard is the end-to-end tool that automates dashboards generation for your AWS infrastructure. How does it work? 1. Dashboard Model Definition To make dashboards flexible and scalable, Dashboards defined by the user using Infinidashboard specify a logical layout with templates, graphs and widgets (common dashboard elements). It enables user-driven dashboard customization by storing templates for common dashboard elements. Infinidashboard extracts a definition of templates, graphs and widgets from the defined dashboards and applies this template when processing a new dashboard. You can define your own templates using the AWS Dashboards Customizable Templates. 2. Dashboard Generation A dashboard can be defined using a set of metadata (template, graphs and widgets) and a set of metadata describing the data source(s). An abstract description of data sources (CRUD operations, security groups and subnets) is shared as a common model that ensures consistent behavior. When processing data from the corresponding data sources, the user-provided templates, graphs and widgets are automatically applied to the specified data and the resulting dashboard is saved into an Amazon S3 location. An Abstract Data Source Model (or ADSM) is a common metadata model that describes a data source. The user creates an ADSM to describe the infrastructure and other metadata associated with the data source (security groups, subnets, RDS instance, Lambda functions, Kinesis stream) that drives the dashboard. 3. Dashboard Security Infinidashboard uses the IAM model to ensure users' data access control and permission policies are properly applied.
- colanderman 5y agoAha nice; good advert for GPT-J. Actually I included a newline in the prompt; I'll edit to make that clear.
- jcims 5y agoThat came across in your original post, I tried to do the same but GPT-J doesn't like whitespace at the end.
- jeffbarr 5y ago> That's hilarious. I think Jeff has an account here. I am 100% sure that I did not write that, and I am 100% sure that this comes across as far too Jeff-like for comfort!
- sillysaurusx 5y agoHahaha. I bet it was a shock to see your name pop up in GPT-3. That feeling when you don't know whether you're famous enough that you were included in the training set and successfully influenced the bot, or the bot simply used a pretty common American name.
- teameat 5y agoYou know, I cannot tell you how many inventions I have been involved with that started out as a few of us just joking around. I bet there is a cool theory or name for this, but sometimes, humor has a way of sort of shaking loose peoples imaginations. hahaha, you never know :) It would be hilarious if you guys end up with something cool from it. Shug.
- pram 5y agoThis is pretty much Kinesis Data Analytics lol