array(2) { ["meta"]=> array(6) { ["name"]=> string(10) "openaq-api" ["license"]=> string(0) "" ["website"]=> string(1) "/" ["page"]=> int(1) ["limit"]=> int(100) ["found"]=> int(0) } ["results"]=> array(0) { } } object(stdClass)#2 (13) { ["coord"]=> object(stdClass)#1 (2) { ["lon"]=> float(125.6081) ["lat"]=> float(9.0125) } ["weather"]=> array(1) { [0]=> object(stdClass)#3 (4) { ["id"]=> int(801) ["main"]=> string(6) "Clouds" ["description"]=> string(10) "few clouds" ["icon"]=> string(3) "02n" } } ["base"]=> string(8) "stations" ["main"]=> object(stdClass)#4 (8) { ["temp"]=> float(77.94) ["feels_like"]=> float(79.81) ["temp_min"]=> float(77.94) ["temp_max"]=> float(77.94) ["pressure"]=> int(1015) ["humidity"]=> int(93) ["sea_level"]=> int(1015) ["grnd_level"]=> int(1013) } ["visibility"]=> int(10000) ["wind"]=> object(stdClass)#5 (3) { ["speed"]=> float(3.67) ["deg"]=> int(106) ["gust"]=> float(7.67) } ["clouds"]=> object(stdClass)#6 (1) { ["all"]=> int(14) } ["dt"]=> int(1711712334) ["sys"]=> object(stdClass)#7 (5) { ["type"]=> int(2) ["id"]=> int(2093017) ["country"]=> string(2) "PH" ["sunrise"]=> int(1711661800) ["sunset"]=> int(1711705664) } ["timezone"]=> int(28800) ["id"]=> int(1705545) ["name"]=> string(11) "Los Angeles" ["cod"]=> int(200) } DrPunchman

Fun With GPT-2 & 3

200 Ads

200 Ads

My first novel, and it's killing me.

I produced the PepTalks Podcast for almost 2 years. During that time I wrote 2-4 ads per episode. What started as my publishing a book that was just the ads, ended up me writing the stories behind the ads as well as a behind the scenes for the show. It's almost 1000 pages and maybe 30% complete.


I needed help pushing past some depression and writer's block. GPT-3 seemed like it might just do the trick.


Then I dove in.

InferKit (Beta)

InferKit (Beta)

Garbage in, garbage out.

My first peek was with InferKit and I wasn't that excited about the results.


It was as if it kinda knew what I was talking about but then continued with a bizarre tangent that ended in an ad for a winery before repeating my text.


My IAOIP group has been working on an international document for innovation, and GPT-2 did a pretty bang-up job knocking out the first draft. Is it possible that I'm too weird for Machines?

Probe

Computer Logic

Possibly the greatest TV series that you never watched is 1988's Probe.


In the first episode (a 2-parter) the city creates an AI that gets obsessed with late night religious TV programming and begins 'sending people to heaven.'


During my research I found that GPT-3 both assumes that it is self-aware, and is obsessed with death.


Maybe my book needs more death?

GPT-3 in Nature

GPT-3 in Nature

What other folk are doing

I began to look at what other people where doing with GPT-3 and found myself beyond the looking glass


Dover uses AI to write job descriptions based on keywords and bullet points, because jog descriptions weren't enough of a mish-mash of regurgitated copy-paste anyway.


Fitness AI gives you medical advice if you are too lazy to call your great aunt for the same information.


AI Dungeon is a text based game that is 100% AI generated. It's the kind of thing I used to build in BasicA with 1000 GOTO statements, but who has time for that now? OK, I probably do. No. No. Stay on target. (I robbed a child, tried to sell him in the city, and got caught attempting to burn down his adopted parent's house. I convinced the dad to burn his OWN house down and turned him in for a reward. This website is horrible magic)


Simplify.so takes complex ideas and breaks them down so a child could understand them. I gave it my 'Roo Cheese' ad and it returned with “So easy to prepare, you don’t even need to peel it” and “The ultimate ALL-IN-ONE kit”.


PhilosopherAI.com generates answers to questions no one really wanted to hear the answers to. So it's good for reference, but... it's not like there is money in philosophy as long as everyone has assholes, as it were.


TagLines.ai is the better version of silmplify.so: “The only kangaroo cheese you will ever eat.”, “Roo cheese is delicious and healthy for you!”, and “Cheese is a food that makes you fat.” and about 10 other decent quips.


storymapr.com and/or ShortlyRead might be exactly what I was looking for.

GPT-3 Making Movies

GPT-3 Making Movies

Like Amazon Prime Does

If you are like me and need a little creative boost while writing, you are in luck. There are several examples of people using this AI to generate movie scripts, and as much as I hate to say it... they are not all bad and they are not all Keaton Patti jokes.


Sejuti Das did a great breakdown on Analytic India Mag where she talks about GPT-3 pretending to be Elon Musk, writing music, and writing a 4-minute screenplay named “Solicitors” that is everything you may expect from a student film.


Netflix has been said to possibly use GPT-2 to write movie descriptions (something no one reads) by BecomingHuman.AI.


But for my money, I'll stick with ForwardDataScience.com and their article on using GPT-2 to generate screenplays. This takes us all the way to ScriptBuddy, a GitHub repository with a scriptwriting AI.