Can an Artificial Intelligence Write A Good Blog Post?

A blog post by an artificial intelligence that has just been taught to write about an artificial intelligence is still far from being a good blog post by any human being.

There are many reasons why a blog post by an artificial intelligence is unlikely to be a good blog post by any human being.

First, the post is still a huge piece of written material, so it will be a big task for the AI to read it all. This is similar to reading a huge, long book, which is a huge task for you as an author.

Second, it is likely that the AI will use its knowledge about writing to create a very bad, misleading, or otherwise nonsensical blog post. In this case, the AI will be writing a blog post about its own stupidity.

Third, even if the blog post is not written by a computer but rather by a human author, the human author will not understand it and will not be able to correct it. This is because the blog post will include very basic, incorrect, or outdated knowledge about writing,

Comments

Is this a trick post by an AI?

Seems so. Generally, Alex Tabarrok posts are way longer and he has his own style which is not this.

The only other explanation for this is that he was kicked in the head by a horse.

No, as that would likely make his post a rant about occupational licensing for AIs...

Respond

Add Comment

Respond

Add Comment

Is this a trick post by an AI?

Answer via:
Probably. But the point is that as I am writing this article, a robot named Sophia is taking over at MIT's Artificial Intelligence Laboratory. I am very pleased.

Respond

Add Comment

Respond

Add Comment

Sure looks like it!

Respond

Add Comment

Oh oh, now I hear the drones above my hou..

Respond

Add Comment

Oh no! They created an artificial hack. Imagine all the soon-to-be unemployed professors, government officials and media types. You maniacs! You blew it up! Ah, damn you! God damn you all to hell!

I hate every ape I see, from chimpan-A to chimpanzee.

Respond

Add Comment

+1, funny

Respond

Add Comment

Respond

Add Comment

Turing Test. First thing I thought of.

It is a post written by a human to appear sparse and awkward while still grammatically correct.

I answer the original question by saying: 1) yes, it's already been done and 2) its irrelevant. Leftist propaganda is easily machine generated. Every speech contains the same empty, conclusory platitudes devoid of facts. And writing has been dubbed down so much that a blog post would be considered a short story by comparison. Young people and exploitive politicians speak in slogans, sound bites, bumper stickers, and tweets. These can easily be machine generated and believed completely by uncritical sheeple.

Lol, and the pivot to blame everything on leftists isn't easily programmable?

Iolololol

Respond

Add Comment

'while still grammatically correct'

Nope - it has the sort of mistake at the very end that only a person utterly unfamiliar with PC grammar checking, ca. 1995, would make.

Respond

Add Comment

Respond

Add Comment

By the way, on human-computer interactive creativity, this is really awesome.

Background on the Demoscene, which was new to me.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

This looks like an airplane post, orange essay by a 12 year old

Respond

Add Comment

gpt2?

Exactly.

Here's a tweetstream by James Ryan, who is interested in the history of computer-generated narrative. He took computer narrative from the old days of symbolic AI and used GPT-2 to extend them. Very interesting.

Respond

Add Comment

Respond

Add Comment

"First, the post is still a huge piece of written material, so it will be a big task for the AI to read it all."

If the AI is doing the writing, why does it care about reading?

"This is similar to reading a huge, long book, which is a huge task for you as an author."

Writing is big task for an author not reading.

Alex, your AI is confused about reading versus writing. Or maybe you were just drunk.

Respond

Add Comment

> Is this a trick post by an AI?

No - not artificial intelligence; it's genuine stupidity instead

Respond

Add Comment

Tabarrok has a great sense of humor. Not to say Cowen doesn't, but I suspect this is one dimension where Tabarrok excels.

Citation needed.

Respond

Add Comment

Tyler gets many genuine laughs in class. Unlike most all other professors who get total kiss ass responses.

Either way, no one reads blogs anymore (with a small handful that is). We have twitter for that.

Au contraire. It's true that most blogs have lost readers but that's because they closed comments; they closed comments primarily because the comments became so offensive. That moved the offensive comments to twitter, where they are welcome and belong. While I may agree with brevity ("If you can't write your idea on a matchbook cover, you don't have a clear idea"), twitter isn't to express ideas but to impress "friends" and offend "enemies". It's a waste of time. Get over it. As for Cowen, I agree he is charming, which is why I read transcripts of his interviews. That way I get over it.

No one reads blogs anymore because free writing platforms weren't generating enough revenue for their big data hosts, so they started steering their traffic to higher RPC websites.

Like Buzzfeed.

Lucky us.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Well done.

Respond

Add Comment

Unfortunately the Tom Friedman Op-Ed Generator site is no longer active. It was pretty fun to play with.

That generator was great. As I told a taxi driver last year in Madrid, it reminded me of Julius Caesar riding on Napoleon's horse.

Respond

Add Comment

Respond

Add Comment

Whether or not this post is created by AI,

It has bad composition and English sentence structure. Redundant.

C-

Respond

Add Comment

A blog comment by an artificial intelligence that has just been taught to write about an artificial intelligence is still far from being a good blog comment by any human being.

There are many reasons why a blog comment by an artificial intelligence is unlikely to be a good blog comment by any human being.

First, the comment can be a huge piece of written material, so it will be a big task for the AI to read it all. This is similar to reading a huge, long book, which is a huge task for you as an author.

Second, it is likely that the AI will use its knowledge about writing to create a very bad, misleading, or otherwise nonsensical blog comment. In this case, the AI will be writing a blog comment about its own stupidity.

Third, even if the blog comment is not written by a computer but rather by a human author, the human author will not understand it and will not be able to correct it. This is because the blog comment will include very basic, incorrect, or outdated knowledge about writing,

+1

AI needs an editor.

Respond

Add Comment

while true
do

sed -e "s/comment/sub-comment/g" tmp.txt

cat tmp.txt

cp tmp.txt ChetterHummin.txt

done

grrr. my bash input and output redirection got eaten up as bad html.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment

Looks like GPT-2: 1.5B. Pretty decent.

Slatestarcodex had a nice post earlier this year with some AI written poetry.

OpenAI now has a bot that can play StarCraft 2 better than 99% of players (although it gets crushed by pros). Interesting times.

Isn't standard hard mode better than 99% of the casual players? Well, it is true hard mode cheats the rules that players have to follow.

Respond

Add Comment

Respond

Add Comment

More relevant Question

Can humans write intelligent comments on the site?

Respond

Add Comment

That is very interesting, Alex. Please tell me more.

You mean, Alexa.

Respond

Add Comment

Respond

Add Comment

wait a second.....

Respond

Add Comment

Where is the final paragraph restating the Main Idea of the first paragraph in slightly different words?

Did the AI (or the AT) reject the need for that? If so, never mind the inanity, the exercise would represent real progress in the pedagogy of grade-school writing.

Respond

Add Comment

The answer to the headline seems to be: "no".

Respond

Add Comment

"There are many reasons why a blog post by an artificial intelligence is unlikely to be a good blog post by any human being."

While this may itself be true, is it not the case that a single solitary reason suffices?

Or: did some AT or AI intend to say instead " . . . is unlikely to be as good as a blog post by a human being"? --or something like that . . . ? --or anything like that?

(If the AI author were as well-trained as any human author, it will have been well-trained to re-read its scribblings prior to posting.)

The end comma in the excerpt was a tell-tale touch.

"There are many reasons why a blog post by an artificial intelligence is unlikely to be a good blog post by any human being." An X created by Y is unlikely to be an X created by Z, unless Z=Y. ...and what is meant by "unlikely". Regardless of what AT's intentions were/are, the post is a waste of space and reader's time. Which makes it a bad post.

Respond

Add Comment

Respond

Add Comment

Étant qu'on a cessé d'enseigner les mathématiques avec la vérification intégrée pour faire la preuve du raisonnement, qu'on a cessé d'enseigner la langue parce l'apprentissage de la grammaire et sa preuve dans l'analyse des dictées, on a tout simplement avec l'avènement de l'informatique décidé depuis aussitôt que le début des années 60' de programmer les humains comme des ordinateurs. Lui préférent l'entreposage des données plutôt que de développer le raisonnement et la créativité. Encore pire avec les algorithms il enlève la possibilité de faire des liens et de chercher des solutions ailleues que dans le sujet énoncé au départ. Alors que plusieurs solutions viennent de d'autres champs de connaissance que ce soit par mimétisme ou transfert. Désolant parce que la plus grande valeur de l'humain c'est la créativité, sa capacité d'adaptation. La nature, la vie est proframmée pour survivre. Si on a une forêt en train de s'acidifier, les feuillus ne pourront plus pousser, la bature va chercher à rétablir l'équilibre en développant des tordeuses d'épinettes pour en réduire le nombre et permettre aux feuillus de s'y implanter de nouceau. Que fait l'humain en idiot il met des produits chimique pour détruire la tordeuse d'épinette. Si on ne prend pas conscience du d3ficit d'intelligence qu'on est en train de créer en empêchant l'intelligence humaine de se développer, en minant la créativité et l'empêchant de raisonner avec son instinct de survie qui trouvera les solutions aux problèmes de demain? Les ordinateurs? Sont-ils supérieur à la nature? Des chercheurs ont d'instinct il y a des sciècles trouvé des hypothèses qu'ils n'avaient pas les moyens de prouver dans ce temps là mais qu'ils savaient d'instinct être l'explication. Et en se basant sur ça ils ont développé des connaissaances. Il faut faire confiance à la vie, la nature est forte et trouve toujours les solutions pour survivre. Il faut éducquer en misant sur la compréhension plutôt que sur l'apprentissage. Comprendre et utiliser des outils où sont entreposé les données. N'enrégistré que les expériences significatives utiles pour survivre. Inutile d'apprendre des recettes par coeur, vaut mieux aporendre à reconnaître la qualité des aliments.

Respond

Add Comment

What if this were the best post

Alex had ever written

And it was an AI composition.

Humiliated by 1's and 0's.

Or, perhaps the bar is set low.

See now this is clever. I like it.

Respond

Add Comment

Respond

Add Comment

Sorry l have written in frebch but been traduce by a stupid program, so you won't probably be able to understand what l have said ; +(

Respond

Add Comment

The solution to badly written AI authored articles is simple and right at hand: simply instantiate more AI's to read the articles, they can even click the ads.

Respond

Add Comment

In its copious spare time, how well or how poorly might an AI reader digest a brief exposition of apophatapataphysics?

http://fictionaut.com/stories/strannikov/illustrated-comments-on-the-apophatapataphysical-metrics-of-cosmic-humor

Respond

Add Comment

This is not a very good Turing test

Respond

Add Comment

The real question here is whether the AI is writing this post with the specific instruction that *people should be able to tell it is an AI.* In other words, is the AI trolling us?

If so, this is proof that an AI can, indeed, write a very good blog post.

Respond

Add Comment

All is clear! Thiago is actually Alex Tabarrok.

Respond

Add Comment

It will be a bad blog post because the AI isn't actually thinking. It's not taking a prompt and trying to reason through ideas. Instead it's using a sophisticated language model to produce words that relate to each other grammatically and thematically. It isn't intelligence in any way that we currently understand the word. It is a statistical model.

Re: "...words that relate to each other grammatically and thematically. It isn't intelligence in any way that we currently understand the word."

Sounds like

A Trump tweet.

Respond

Add Comment

Respond

Add Comment

Cute. But I'm struck by how the repetitiveness and unexpected orthogonal shifts in the narrative direction are reminiscent of some autistic people.

Respond

Add Comment

I dunno. An internet - where AI writes blog posts, which are read by other AI, and then argued about by bots in the comments sections - that strikes me as an improvement.

Respond

Add Comment

On the Internet, no one knows you're an AI.

Respond

Add Comment

Read this subreddit, where AIs trained on the content of other popular subreddits create and comment on posts. The ones written and commented by a single AI are interesting, the ones labeled "Mixed" are participated in by all AIs and are comedy gold. At least for this implementation of AI, there's good understanding of syntax and little understanding of meaning.

https://old.reddit.com/r/SubSimulatorGPT2/

Respond

Add Comment

Not only does Alex not explicitly write the true meaning of this text, he didn’t explicitly write the text at all! This is the ULTIMATE form of Straussianism!

Respond

Add Comment

Your circle is circular.

Respond

Add Comment

Robert Laughlin won a Nobel prize for being one of the first talented physicists to study with real passion what is called "emergent phenomena" in physics.

An AI post imitating a real post is not an emergent phenomenon, it is an artifact.

An emergent phenomenon is this (well, it is the last stage of this set-up: bear with me)

- let's say I really like Tabarrok posts, and I am sad that there are too few of them.
If I want to read more, I have two sets of choices - (a) I can hope that Tabarrok posts more often, I can send money to Tabarrok as an incentive to post more often, I can do similar positive things or
(b) I can spend a couple hundred bucks, spent in the way only connoisseurs know how to spend money, on an "AI" filtering system that will go through billions of internet-hosted comments every second and will, in a few minutes, deposit in my inbox a few dozen out of the billions that are "the sort of thing Tabarrok might have said on a given subject had Tabarrok actually said something on the given subject".

Bear with me, that is not the emergent physics of this algorithm.

What is emergent is this ----- if I really really want to read more Tabarrok posts, I will pay a few hundred people to go through the motions I described in (b) and I will pay them to rewrite, even closer to Tabarrok style, those non-Tabarrok posts that turned up in their in-box.

Real AI performance will occur when that last task can be performed by non-sentient AI. And one step past that is the emergent step.

Somewhere around 2025, someone who wants to understand Shakespeare will be able to read thousands of Shakespearean lines
describing their old Facebook posts

more importantly, describing in articulate and genuinely well-thought-out novo-Shakesperean language what was real about those Facebook posts

2025.

And nobody is gonna care that it is novo-Shakesepearean and not real Shakespeare, no more than anybody said I ain't gonna buy a Louis Armstrong album because there were probably better musicians who lived before he lived

Louis sold a bucket load of albums every day of his adult life

(actually, I have friends in Silicon Valley, and I prefer this does not happen until 2030 at the latest ----- and my friends in Silicon Valley actually listen to me once every blue moon or so, so maybe 2025 is not accurate, and I should say 2030 or 2035 or 2040. Anyway, this is not my world, it is not yours, although there is this:

"King Solomon, in the course of writing the Book of Proverbs which to this day bears his name (not that it does him any good), often reflected on the duty decent and good and brave men have to warn the foolish from putting too much trust in the clever. Only God knows our hearts and it is only by keeping what we always knew for a certainty - that God loves us, and created us, and wants us to be happy - it is only by keeping what we always knew for a certainty in our hearts that we can avoid the stupid temptation to follow the paths of evildoers."

Thanks for reading.

It is no small thing to be a friend to a creature who never had a friend in this world

)

Respond

Add Comment

Respond

Add Comment

Sorry for the long comment, I get a little too enthusiastic when there is a post here about something I probably know more about than almost anybody else who has ever lived

To be fair sometimes my comments are short ----

PROVERBS 8 !!!!

I remember.

Respond

Add Comment

Respond

Add Comment

Respond

Add Comment