Ever since the head-turning implications of auto-text generator GPT-3 became well known, many in the AI-generated writing community have been holding their collective breath.

Admittedly imperfect, the experimental auto-writing tool is nevertheless capable of astounding feats, leaving some of the top computer scientists in the world clamoring for a test-drive.

“The hype is real,” observes Kelsey Piper, a writer for Vox.com who personally tried out GPT-3. “It has its shortcomings, but make no mistake: GPT-3 represents a tremendous leap for AI.”

Adds Lou Kerner, founder, The Social Internet Fund: “GPT-3 is amazing. It’s scary. It’s exhilarating.”

For starters, GPT-3 can generate fake news articles that are very tough to tell apart from the real thing – simply by taking a phrase of input from a human operator and running with it, according to GPT-3’s creators.

Full-length emails can be conjured by the tool after you type in just a few bullet-points to get it started, according to OthersideAI.

And GPT-3 can generate blog posts that seem so authentic, one of them rocketed to number one in popularity on Hacker News – a collaborative news aggregator fed by some of the Web’s most discerning intellects.

Still other startling feats from GPT-3 include its ability to auto-translate everyday language into legalese, quickly generate headlines and tweets and forge any number of works of creative fiction and poetry.

There’s even a new GPT-3-based tool on the Web you can use to auto-generate your own cyberpunk, apocalyptic, zombie or other fantasy simply by typing in a few words at a time.

One of the reasons GPT-3 is so incredibly powerful – and smart — is that it has a powerhouse, well-funded research group – OpenAI – behind it. Founded in 2015, OpenAI became a ‘capped’ for profit organization in 2019, reaping one billion in funding from Microsoft the same year.

Another driver behind its formidable prowess is the software’s penchant for voracious reading: GPT-3 has ‘read’ and ‘ingested’ a mind-boggling amount of published Web content.

Specifically, the tool now has word-for-word access to the lion share of Web pages – and all the data stored on those Web pages — published on the Internet from the beginning of 2016 through the close of 2019, according to Tiernan Ray, a writer for ZDNet.

In addition, that data, which was tapped from Web monitoring organization Common Crawl, was supplemented with a number of additional datasets – including entire collections of digitized books, according to Tiernan.

Armed with a fair approximation of the knowledge of the ages, GPT-3 is able to instantly auto-generate text by grabbing a few words from you and using its AI engine — driven by predictive statistics — to ‘guess’ the kind of text you want.

Interestingly, like many computer applications, GPT-3 runs on a classic ‘garbage in, garbage out’ format. Enter a few vague words into its system and you’ll likely retrieve a vague facsimile of what you’re really looking for.

Enter some precisely chosen words – or precisely chosen longer phrases — into GPT-3, and you’ll most likely get something back that is much closer — or even exactly — what you really want.

Even so, one of the most astonishing aspects of the experimental tool is its ability to guess very closely just what you’re looking for – even if you’re only entering a few words, and even if you have not familiarized GPT-3 with the specific kind of text you want.

“GPT-3 doesn’t need much extra training,” observes Farhad Manjoo, an opinion columnist for The New York Times.

“Give GPT-3 a natural-language prompt — “I hereby resign from Dunder-Mifflin” or “Dear John, I’m leaving you” — and the software will fill in the rest with text that is eerily close to what a human would produce,” Manjoo adds.

Kerry Harrison, who has personally test-driven GPT-3, agrees: “Even more impressive is the fact it can perform specific tasks without any tuning. You can task it to write code, generate poetry, write articles or engage in a Q&A and it’ll handle them all.”

Of course, as experimental software, GPT-3 needs some tweaking before it can be commercialized.

Sure, the tool can generate incredibly accomplished emails, fiction, blog posts and more. But its performance is not completely reliable: Sometimes it can make little mistakes during content generation.

And sometimes it can make some really big mistakes.

Plus, while content generated by the tool is generally ‘just about’ what you’re looking for, a human being will often still need to step-in and edit GPT-3’s work before it can be posted, published, sent as an email or the like.

Still another shortcoming: GPT-3 is notorious for veering off into the nonsensical when it attempts to create long-form fiction and other long-form writing.

In fact, OpenAI CEO Sam Altman is the first to admit that his organization’s experimental AI needs significant polishing before commercial applications can start cropping up.

“It’s impressive – thanks for the nice compliments! –,” Altman tweeted after seeing computer aficionados across the globe swoon over the introduction of GPT-3.

“But it still has serious weaknesses and sometimes makes very silly mistakes,” Altman added.

“AI is going to change the world,” Altman observes. “But GPT-3 is just a very early glimpse. We have a lot still to figure out.”

Part of getting from here-to-AI-nirvana will be learning how to cajole GPT-3 to generate a desired piece of text in just the right way.

For example: Ask GPT-3 to write a poem by entering a few phrases and you’ll most like get back workman-like verse.

In contrast, feed GPT-3 the complete works of a specific poet and you’ll probably be treated to a creation mimicking the style of that poet, according to Tiernan.

Adds James McDermott, a lecturer in computer science at National University of Ireland, Galway: GPT-3 “researchers are becoming horse-whisperers, figuring out what to say to get it into the right frame of mind for some task.”

Vox’s Piper agrees: “If you prompt GPT-3 to write you a story with a prompt like ‘here is a short story,’ it will write a distinctly mediocre story.

“If you instead prompt it with ‘here is an award-winning short story,’ it will write a better one.”

In the meantime, as we await the looming perfection of GPT-3 and the countless commercial applications that will follow — including auto-generated news articles, Web content, fiction, scripts, advertising and more – some have begun to fear anew that the days of humanity’s supreme intellect are numbered.

*”GPT-3 anxiety is based on the possibility that what separates us from other species and what we think of as the pinnacle of human intelligence — namely our linguistic capacities — could, in principle” someday be replicated by machines, observes Carlos Montemayor, a professor of philosophy at San Francisco State University.

*Special Feature: Company Reports That Write Themselves

Share a Link:  Please consider sharing a link to https://RobotWritersAI.com from your blog, social media post, publication or emails. More links leading to RobotWritersAI.com helps everyone interested in AI-generated writing.

–Joe Dysart is editor of RobotWritersAI.com and a tech journalist with 20+ years experience. His work has appeared in 150+ publications, including The New York Times and the Financial Times of London.