Hacker News new | ask | show | jobs
by teameat 1813 days ago
This reminds me of a practical joke a co-worker and I pulled some years back, we made up a new project called the LMNOP project (just letters out of the alphabet). We used a program i wrote based on the old foggy where whenever we would see a good BS line from an email or document we put it in the BS generator DB. We had the program spit out several pages of total BS and some fake but impressive looking diagrams made up of a bunch of boxes with random letters as labels and lines between them. We did put some actual sentences of our own in the beginning and end to hype it. Then we set the trap and left it on the copy machine half copied and walked off. Within a couple of days, bootleg copies were flying around and someone scanned it in and it was being emailed around. A lot of people in our department and even other departments were asking about it. If someone asked us we would say "sorry I can't discuss it" in a very mysterious way. Then we started hearing other people saying that same phrase when they were asked about it. The whole things started to take on a life of it's own, we joked about how it was like a Dilbert come true and there would be an entire team headed by a director with a budget running a program that was fake. Of course that never happened, our director called me in and had a copy of the fake document, and said "this looks like something you may have had something to do with ?". I laughed and said just having a bit of fun. She sent out a blast email telling everyone that it was a practical joke gone viral and for everyone to ignore it. Even so, for a long time people kept coming up with conspiracy theories that was a cover story. It just shows how people like believing stupid stuff.
5 comments

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;

That'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.

> 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!

Hahaha. 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.

You 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.

Aha nice; good advert for GPT-J. Actually I included a newline in the prompt; I'll edit to make that clear.
That came across in your original post, I tried to do the same but GPT-J doesn't like whitespace at the end.
So... now I'm wondering whether this actually should be a thing. That sounds pretty slick.
This is pretty much Kinesis Data Analytics lol
> this looks like something you may have had something to do with ?". I laughed and said just having a bit of fun.

Surely this would've been the best time to reply with "Sorry, I can't discuss it"

haha, I'll put that on my list of quips I wish I had thought to say at the time.
Catch-22 vibes
Reminds me of how a few places I’ve worked momentarily embraced Tableu with open arms after seeing a US revenue heat map built by dragging and dropping (carefully groomed data).
Ya, we had a VP who somehow some sales guy got to and showed him a "magical" Network Management system by showing a carefully staged demo. It took forever to talk him down.
The perfect demo.
The earlies version I saw of this was way back in my Bell Lab days, there was a program called foggy (I think it actually came with early versions of UNIX), I actually still have the original source someplace I think. It was very basic, lead ins, transitions, objects then randomness. Mine was based on that with some tweaks. But these new ones out there are AMAZING.