The latest technology and digital news on the web

Human-centric AI news and analysis

A belled GPT-3-generated blog shows AI still can’t imitate human writing

Last week, many tech publications broke news about a blog generated by bogus intelligence that fooled bags of users and landed on top of the Hacker News forum. GPT-3, the massive accent model developed by AI analysis lab OpenAI, had accounting the articles.

Since its absolution in July, GPT-3 has caused a lot of action in the AI community. Developers who have accustomed early access to the accent model have used to do many absorbing things, assuming just how far AI analysis has come.

But like many other developments in AI, there’s also a lot of hype and confounding surrounding GPT-3, and many of the belief appear about it adulterate its capabilities. The blog accounting by GPT-3 resurfaced worries about fake news onslaughts, robots artful humans, and technological unemployment, which have become the authentication of AI reporting.

I absitively to take a deep look at the blog and the action surrounding it, and my allegation were troubling. But the problems I found were mostly with humans, not GPT-3.

The AI-generated blog

screenshot of adolos blog
Screenshot of Adolos, a blog accounting by GPT-3

In case you haven’t read the stories, a computer science apprentice at the University of California, Berkeley, set up a blog on Substack under the pseudonym Adolos. OpenAI has currently made GPT-3 accessible to a bound admirers of developers, and Liam Porr, the student, was not one of them. So he asked a Ph.D. apprentice who already had access to the AI to run his queries on GPT-3.

Basically, Porr gave a banderole and intro for the post, and GPT-3 alternate a full article. He chose the best of several outputs of the AI model and copy-pasted it into his blog with very little editing.

The first post, titled, “Feeling unproductive? Maybe you should stop overthinking” accomplished the number one spot on Hacker News with nearly 200 upvotes and more than 70 comments. In one week, the blog accomplished 26,000 views and acquired 60 subscribers. According to Porr, very few people had acicular out that the blog might have been accounting by AI.

Porr ended the experiment with a acknowledgment and some belief on how GPT-3 could change the future of writing.

Poor AI reporting

Naturally, such a ambience is an adorable accountable for amazing articles. And the media did not disappoint. I looked at the advertisement of several acclaimed tech publications. The one thing they all used in their annual was the term “fake blog.”

fake blog headlines
Tech media referred to the AI-generated blog as “fake blog.”

The word “fake” is vague to start with. We use it about to refer to affected accessories (fake Nike shoes) or bogus (a fake passport). It can also mean affectation (faking surprise) or clothing and absorb some form of cheat or bamboozlement (fake news).

For instance, during the runup to the presidential elections, a group of Macedonian youth set up what looked like real American news websites and used them to spread fake news accessories with false advice and amazing headlines. The accessories tricked users on social media to click on their links and advance them, breeding acquirement for the sites’ owners and wreaking havoc during the U.S. elections.

Looking at Porr’s blog, I couldn’t see how the analogue “fake” could apply to the blog. The author was not overextension misinformation. He wasn’t trying to access public assessment by giving a false anecdotal of events. And he never mentioned the word “fake” in his own annual of the events.

The author used the penname Adolos, which is acutely a pseudonym or at the very least an abridged name. Using a pen name is a known and accustomed convenance among bloggers. There’s annihilation wrong with it as long as you’re not using it for ambiguous motives or to cause harm to other people. So, I wouldn’t count that as an altercation for calling the blog fake.

Also, the fact that an AI helped write the accessories didn’t make them fake. It did make them altered from human writing, but not fake. I think the term “AI-written” or “AI-generated” would have been more precise.

But then again, the term “fake” is also very subjective. For instance, there isn’t a accord on what fake news is today. Conceivably the writers of those news belief have their own affidavit for calling the AI-generated blog fake. But in the same vein, I could call their belief “fake” for ambiguous their admirers about the capabilities of GPT-3.

But one thing is for sure. Putting “fake blog” in the title will accomplish a lot of clicks and acquirement for ad-driven media outlets, abnormally as the acuteness is at an best high before the 2020 U.S. presidential elections.

The Hacker News post

The media used the AI-generated blog’s acceptance on tech-focused Hacker News forum as a admeasurement that GPT-3 had managed to fool readers that a human had accounting the posts.

Hacker News post on AI-generated blog accounting by GPT-3
A post accounting by GPT-3 made it to the top of the Hacker News forum

The post has indeed accustomed 198 points and 71 comments. “What most commenters didn’t realize: The post was generated absolutely by bogus intelligence,” Business Insider wrote.

But a closer look at altercation paints a clearer annual of why the post performed very well. There are 22 animadversion accoutrement in the Hacker News discussion. Only one of them is an approval of the points raised in the article. Most of the comments focus on other users’ angle and acquaintance on ambidextrous with unproductivity. Some of them had debated the title of the commodity (which was accounting by a human, by the way).

Basically, this hints that, rather than being absorbed in the AI-generated article, the association had found the topic to be absorbing and altercation worthy. In fact, I think the points raised in the comments were much more absorbing than the commodity itself.

Regretfully, none of the media outlets accoutrement the story took care to look into this. Instead they (and Porr himself) accent one animadversion where user had voiced their suspicion about the commodity being accounting by GPT-3, which was downvoted by others.

hacker news downvotes animadversion on GPT-3
Users on Hacker News downvoted a animadversion that declared the commodity was accounting by GPT-3

I think this is pretty natural. While the commodity was accounting by an AI, the altercation was purely human, and some people were apparently afterward it with absorption (more the reason to upvote the post itself and bring more people into the discussion), which comes with some expectations of participants to remain affable and on-topic.

The blog stats

According to Porr, the blog accustomed 26,000 views and 60 subscribers in one week. Again, the media picked this up as proof that AI had fooled people into cerebration a human had accounting the blog.

Here’s an extract from The Verge: “The post went viral in a matter of a few hours, Porr said, and the blog had more than 26,000 visitors. He wrote that only one person accomplished out to ask if the post was AI-generated, although several commenters did guess GPT-3 was the author.”

But 26,000 views doesn’t mean 26,000 people enjoyed the article. It only means that many people found the title of the accessories arresting enough (do I need to remind you the titles were accounting by a human?) to click on them.

I would also want to know more before I would use the view stats as a admeasurement of the blog’s popularity. Were the view stats broadcast across all blog posts or did they mostly belong to the one accustomed post that made it to the top of Hacker News? How many return users did the blog have? How broadcast were the cartage channels of the blog? How affianced were the subscribers with the blog’s posts? What is the blog’s bounce rate? Answering these questions would give us a better annual of the amoebic virality of the blog and how well the AI had managed to argue its readers its accessories were genuine.

Hacker News is a top-10k website on, which means it receives millions of visitors per month. I doubtable that the views of the blog spiked when that one post that made it to the top of forum, and then plateaued at a very low daily rate when it alone off the chart. The news advantage in the recent week has apparently given it addition boost in traffic.

I did a quick search for “” on Cheep to see how many users were administration the blog’s content. Recent shares were caused by the media hype around GPT-3 having accounting the blog and most users are discussing how acceptable the AI autograph is. But if you scroll down to mid-July, when the abundance commodity was published, the abundance of shares reduced, and it mostly included bot accounts that adviser top posts on Hacker News. As the commodity started going up the chart in Hacker News, a few other users also shared it.

Twitter bots share GPT-3 blog post
Many of the Cheep users administration the AI-generated blog were bots that adviser Hacker News top posts

The other accessories in the blog accustomed very few shares, which speaks much about the blog’s popularity.

According to Porr, aside from a few users on Hacker News, only one person accomplished out to ask whether the blog was accounting by GPT-3. This was addition one of the key highlights of the accessories accounting about the AI-written blog.

Again, I think there’s a confounding of the stats here. One user cogent doubts about the blog being accounting by AI doesn’t mean others didn’t have such suspicions. Also, the topic of the blog was artistic thinking, which means many of the people who read it didn’t necessarily know about GPT-3 and advances in natural accent processing.

There’s a likely chance that a lot of people got balked by the poor and inconsistent autograph and left the site after adorable back. And a few more people might have seen the admonition signs of AI autograph but didn’t bother to animadversion on an bearding blog that was just set up a week ago.

To give more context: People are more likely to point out mistakes if they see it in a acclaimed source (say  or ). But when you see a poor autograph domain-less blog with bad writing, you’ll just abolish it as one of the millions of other bad websites that exist.

How well does GPT-3 write?

To added investigate, I read a few of the accessories on the blog, starting with the one that became very popular, “Feeling unproductive. Maybe you should stop overthinking.”

It’s not best writing, absolutely not commodity a able writer would deliver. There was a lot of repetition. I had to re-read some of the sentences to grasp the meaning.

But even though my mind was primed to look for signs of bogus writing, I had to admit that it stayed on topic, and it didn’t have the ambagious references found in other AI writing. It had bendability and read more like an commodity accounting by a non-professional writer. It shows how far AI has come in spitting out articular text.

In fact, it was accounting well enough that some users became apprehensive about AI having generated the text. “Now, if you spin this added I could come and guess that the  experiment here is that you absolutely wrote the ‘overthinking’ commodity yourself, are now claiming that GPT-3 did it and keep on watching the accessible debate about it,” one user wrote after Porr made the revelation.

Was the blog accounting by GPT-3 or by a human

So, was this really GPT-3 autograph articular text or a huge publicity stunt? Did GPT-3 manage to nicely stitch calm parts of its training data? Was there more than a little human help involved?

At this point, I can neither affirm nor reject cabal theories. But as I arrested the other accessories in the blog, the affection of the autograph was visibly inferior to that of the overthinking post.

For instance, in this article, the AI starts with ambidextrous with plateaus when autograph new posts. Then he talks about a friend who had shared acquaintance about hurdles in Marines bootcamp. Added in the article, the author speaks about his own time in Marines bootcamp and then moves on to the business world. Although there’s a sort of logic involved, the arrangement of events is more than a bit confusing.

There are also signs of human manipulation. For instance, in the same blog post, one of the paragraphs starts with: “Since I’ve started this blog I’ve affected one plateau after another.” When spinning out articles, GPT-3 knows annihilation about the medium where it will be appear or the antecedent accessories appear there. Porr would have to be acutely lucky for the AI to have about generated that sequence.

The only way we can find out the truth is to accomplish some reproducibility experiments. Porr would have to acknowledge full capacity of how he used GPT-3. This includes the agreement of the randomness constant and the acknowledgment length. We would also have to know how much of the intro for each commodity was accounting by Porr himself. Then addition who has access to GPT-3 can run the same queries in the AI and check whether the output (or its quality) matches the accessories on the Adolos blog.

What is the impact of GPT-3?

In his final blog post, Porr declared his observations, including the shortcomings of GPT-3: “If you read some of the agreeable I made, you may not be assertive about its quality. Indeed, there are traces of illogic, adversity with blockage on topic, issues with repetition, etc.”

This is why he chose abundance and self-help as the topic of his blog posts. “GPT-3 is great at creating admirable accent that touches emotion, not hard logic and rational thinking,” he writes.

If you look at the articles, they mostly read like claimed experience. There’s no fact-based logic involved, which would make it easier to hide the inconsistencies and hard to debate the accuracy of the claims.

Porr believes GPT-3 can become a autograph tool and help writers become more advantageous and save media companies millions of dollars by acid staff. Alternatively, according to Porr, GPT-3 will give rise to a new breed of “fast and lean” media companies. These organizations use AI to create vast amounts of accessories and small teams that only make the final edits to fix analytic mistakes and inconsistencies.

After a fashion, he’s right. There’s a lot of poor agreeable out there. Many of the things you read on the web are spinoffs of other articles. There’s too much cheap appropriation and too little aboriginal content. GPT-3 might be able to automate all those tasks and put many “content writers” out of work.

But this only shows how poor human autograph has become, not how good AI autograph is. People are autograph accessories for search engines, for social media content-ranking algorithms. As we have come to rely on algorithms to curate our content, our own autograph has become optimized for those algorithms. And that is commodity that can be automated. GPT-3 or some other AI might enable agreeable farms and online media to fill social media feeds and search engine after-effects pages after the need for human writers.

But it won’t necessarily lead to an access in revenue, as one user acicular out in the comments area of Porr’s final blog, and can have the about-face effect.

effects of GPT-3 on autograph and media

What will the impact be? Overall, there will be some adjustments, but I don’t think people will stop annual online agreeable or lose trust in accounting content. In contrast, it might lead to more acknowledgment for human creativity.

The rise of AI-generated accessories might cause a shift in the way people find agreeable online. For instance, as the affection of search after-effects and social media feeds decreases, the work of human curators who find affection accessories and share them with their admirers will become more valuable. An archetype is the newsletters of business analysts Benedict Evans and Shelly Palmer, who consistently share read-worthy accessories from around the web. The assimilation of the web might in fact create new jobs for people who can abstracted the wheat from the chaff.

AI autograph is real, but it won’t alter humans

First, hats off to Liam Porr for the absorbing experiment. Although I do not agree with his conclusions, I think his work has accent some axiological problems we face.

The media need to change how they cover advances in AI. They should take a deeper look at things before autograph amazing accessories and creating ambiguous hype about accent models.

As I’m autograph these last lines, I’m absorption on the process, from the moment I had the idea for this article, to the investigation, my antecedent acquaintance with the hype surrounding AI, and aggregate else that led to what you’re annual now. Could all of this be automatic by a apparatus that looks for patterns in numbers? I don’t think so.

I think GPT-3 will have a role in writing. Give it a query and it will find commodity accordant in the terabytes of data it has analyzed. That can help you find new admonition for your writing, and it is how AI is being used in other fields. But for the time being, the human artistic action will be beyond GPT-3 or its successor(s).

Some might argue that the same was said for chess and Go, but both proved to be problems that could be solved through brute-force math. They might be right, but accent is conceivably the most complicated action of the human mind and probably one of the last areas AI will conquer. So there will be many other signs before we see AI that can imitate human writing.

To fellow human writers, don’t worry about being replaced with AI, unless you’re autograph for the robots (search engines, social media networks, etc.). Things might change when (if) addition manages to create human-level AI, but it apparently won’t happen in our lifetimes.

And to the robots annual this article, no hard feelings.

This commodity was originally appear by Ben Dickson on TechTalks, a advertisement that examines trends in technology, how they affect the way we live and do business, and the problems they solve. But we also altercate the evil side of technology, the darker implications of new tech and what we need to look out for. You can read the aboriginal commodity here.

Appear September 1, 2020 — 09:27 UTC

Hottest related news

No articles found on this category.