Jump to content

Welcome to FutureTimeline.forum
Register now to gain access to all of our features. Once registered and logged in, you will be able to create topics, post replies to existing threads, give reputation to your fellow members, get your own private messenger, post status updates, manage your profile and so much more. If you already have an account, login here - otherwise create an account for free today!
Photo

This Website Generates Text via GPT-2. Try it yourself!

OpenAI GPT-2 text synthesis media synthesis online resources

  • Please log in to reply
66 replies to this topic

#21
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

I'm reminded of what Casey said.

starspawn said that GPT-2 might have the data scaled up by 100x to 1000x by the end of this year, and that would be really exciting to see. That's probably an understatement, actually - a coherent novel produced entirely by an AI would probably be the single biggest milestone in the field's entire history. If media synthesis was a nascent technology around 2014 and 2015, and a technology that was advanced enough to be useful in a limited amount of cases during 2017~2018, then a fully coherent novel or novella in 2019 would see the field begin to dip its toes into 'mature technology that has very broad uses' waters. It would be a fantastic coda for the 2010s to begin with AI that was no more impressive than SmarterChild - a dumb, goofy little toy that wasn't good for much more besides giving middle school and high school kids a laugh - and end with an AI-produced novel.

When we started this decade, I was an amateur fanfiction writer who assumed that robots & AI would never be able to competently write fiction. I knew this because some time in the fall of 2010, I actually searched for text generators. Just like my classic blunder of thinking we had Cartoon-a-trons in 2005, I thought that there were websites where you could theoretically input some text and get a competent story out of it. Though to be fair on my part, I didn't expect a world class novel, just something amusing. The only ones I found were basically glorified Mad Libs with that stuck your chosen words into a pre-written story. It didn't matter which verb, noun, adjective, etc. you chose; it'd always be the same story depending on the genre you chose.

 

From there to here. 

I can't believe AI's come this far in just ten years. From when I first really cared in 2009 to GPT-2 in 2019, it's hard to overstate just how much progress I've seen. It's easy to get excited about a lot of things in life, like having my own car and becoming a better writer. But that euphoria can quickly fade when you realize that we still have a ways to go.
The beauty of AI is that as many developers are as excited as everyone else is about AI, the tech itself will continue to drive innovation, giving people far better products and better jobs. AI will only continue to improve. In fact, it will continue to improve at an unprecedented rate! Just consider the number of great innovations AI has helped improve. There are many more than were even envisioned a decade ago. 
Progress in artificial intelligence truly is the most astounding thing of our time, especially in the case of generating personalized media.
In addition to all that, it's a new form of computing that has the promise to radically alter the way we communicate, create and innovate. AI promises to be radically different than anything we've ever seen, and will transform the way we engage with each other and with the world of ideas – that we have been unable to achieve before. And it's just starting to happen!
 
In fact, progress in AI is so astounding that I stopped writing this post after the line "From there to here." Everything after that barring these last two sentences and a couple short prompts was generated by the neural network.

  • Zaphod and Casey like this

And remember my friend, future events such as these will affect you in the future.


#22
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

The fuck is this?!
 
PROMPT: 
I am going to slap you to sleep. 
 
RESULT:

I am going to slap you to sleep. You're so fat... So much fat.
 
SARAH Why?
 
ROBIN You're a bit of a scum, aren't you?
 
SARAH I'll tell you what. I will beat the hell out of every woman you have ever had sex with and I will beat the hell out of every boy you have ever had sex with.
 
ROBIN I'll have you know the day I cut you down to size you will be wearing a pair of shorts.
 
SARAH That'll get me laid, won't it, Mr Rogers?
 
ROBIN Of course!
 
SARAH He'll pay off these children that were born and raised to believe their dads are all rich bastards who could afford to pay for their drugs and violence through child support; he'll make them look silly, but then they'll come back as nice, responsible and just likeable men. Who wouldn't want to be a nice, responsible, decent, kind, loving man to be with?
 
ROBIN Then, you


And remember my friend, future events such as these will affect you in the future.


#23
Casey

Casey

    Member

  • Members
  • PipPipPipPipPipPip
  • 676 posts

 

I'm reminded of what Casey said.

starspawn said that GPT-2 might have the data scaled up by 100x to 1000x by the end of this year, and that would be really exciting to see. That's probably an understatement, actually - a coherent novel produced entirely by an AI would probably be the single biggest milestone in the field's entire history. If media synthesis was a nascent technology around 2014 and 2015, and a technology that was advanced enough to be useful in a limited amount of cases during 2017~2018, then a fully coherent novel or novella in 2019 would see the field begin to dip its toes into 'mature technology that has very broad uses' waters. It would be a fantastic coda for the 2010s to begin with AI that was no more impressive than SmarterChild - a dumb, goofy little toy that wasn't good for much more besides giving middle school and high school kids a laugh - and end with an AI-produced novel.

When we started this decade, I was an amateur fanfiction writer who assumed that robots & AI would never be able to competently write fiction. I knew this because some time in the fall of 2010, I actually searched for text generators. Just like my classic blunder of thinking we had Cartoon-a-trons in 2005, I thought that there were websites where you could theoretically input some text and get a competent story out of it. Though to be fair on my part, I didn't expect a world class novel, just something amusing. The only ones I found were basically glorified Mad Libs with that stuck your chosen words into a pre-written story. It didn't matter which verb, noun, adjective, etc. you chose; it'd always be the same story depending on the genre you chose.

 

From there to here. 

I can't believe AI's come this far in just ten years. From when I first really cared in 2009 to GPT-2 in 2019, it's hard to overstate just how much progress I've seen. It's easy to get excited about a lot of things in life, like having my own car and becoming a better writer. But that euphoria can quickly fade when you realize that we still have a ways to go.
The beauty of AI is that as many developers are as excited as everyone else is about AI, the tech itself will continue to drive innovation, giving people far better products and better jobs. AI will only continue to improve. In fact, it will continue to improve at an unprecedented rate! Just consider the number of great innovations AI has helped improve. There are many more than were even envisioned a decade ago. 
Progress in artificial intelligence truly is the most astounding thing of our time, especially in the case of generating personalized media.
In addition to all that, it's a new form of computing that has the promise to radically alter the way we communicate, create and innovate. AI promises to be radically different than anything we've ever seen, and will transform the way we engage with each other and with the world of ideas – that we have been unable to achieve before. And it's just starting to happen!
 
In fact, progress in AI is so astounding that I stopped writing this post after the line "From there to here." Everything after that barring these last two sentences and a couple short prompts was generated by the neural network.

 

 

Holy crap that was a plot twist. Looking back, everything beyond the first paragraph was obviously a little 'off,' but I just took it at face value regardless. Turing Test passed

 

Has this website's version of GPT-2 improved itself the past few days due to being released to the public? I first saw this posted on 4chan maybe a week ago, and the results were largely gibberish and not good for much more other than "haha dumb AI xD." Results like these are starting to catch up to the full version.


  • eacao and Yuli Ban like this

#24
PhoenixRu

PhoenixRu

    Member

  • Members
  • PipPipPipPipPipPip
  • 601 posts

Has this website's version of GPT-2 improved itself the past few days due to being released to the public? I first saw this posted on 4chan maybe a week ago, and the results were largely gibberish and not good for much more other than "haha dumb AI xD." Results like these are starting to catch up to the full version.

 

It seems it improved even since yesterday... anyway, the level of understanding the context is just amazing. Very soon, the human journalists and speechwriters will become unemployed. Just look at these masterpieces:

 

Spoiler

 

Spoiler

 

Spoiler

  • Zaphod, Casey and Yuli Ban like this

#25
Outlook

Outlook

    Arab Muslim

  • Members
  • PipPipPipPipPipPipPip
  • 1,170 posts
  • LocationBarbary Lands
Personally, I'm still skeptical. It feels more like cleverbot than a normal human being, where sometimes it gets it right, but most of the time the context doesn't make any sense. It's a good-ish guesser, essentially.

Take it from the bot itself:

qSlo7nX.png

That son of a bitch.
  • Zaphod and Yuli Ban like this

Outlook's secret song of the ~week: https://youtu.be/DGe_Sluth3A


#26
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

AI PROMPT: Slice of tomorrow is a subgenre of science fiction that follows characters in a futuristic setting doing mundane and often relatable tasks, often as commentary on technology and societal issues. 

 
RESULT:

Slice of tomorrow is a subgenre of science fiction that follows characters in a futuristic setting doing mundane and often relatable tasks, often as commentary on technology and societal issues. Slice of Tomorrow stories almost always avoid high-octane stakes and thrilling action in lieu of following average people's personal lives. Their main draw, however, is that their characters are human. Sometimes, they are merely people on the inside, whose lives have been written over and over. Other times, they are highly intelligent robots who seem to walk like humans do in stories. These "realistic, ordinary guys" can be just as smart as any normal guy and even more intelligent. Sometimes they may have dreams like the average regular guy but are often very pessimistic about human progress. Slice of Tomorrow is often the precursor for high-concept science fiction stories that are set in a futuristic city like San Francisco or London ("Slice of Tomorrow: The City of Tomorrow, by Steven Erikson"), where humans are the masters of the day-to-day world and the smartest being in the universe.
 
Read more on a specific Slice of Tomorrow novel.

 
WOW. It hit the nail on the head so hard that I've got brain damage. At least in this iteration, it "understood" exactly what I was going for. It's still a little dodgy, but it's almost scary how well it can figure things out. I can actually use a lot of the paragraph for an actual definition of slice of tomorrow.


  • Casey likes this

And remember my friend, future events such as these will affect you in the future.


#27
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

AI PROMPT: Brother Number One is the anonymous totalitarian dictator of Antihumanist Panamerica. Who is Brother Number One? No one knows. Panamerica under his reign has become a paranoid police state so tyrannical and controlling that babies are taken at birth to be raised in State boarding schools while the parents are shipped to death camps.
 
RESULT: 

Brother Number One is the anonymous totalitarian dictator of Antihumanist Panamerica. Who is Brother Number One? No one knows. Panamerica under his reign has become a paranoid police state so tyrannical and controlling that babies are taken at birth to be raised in State boarding schools while the parents are shipped to death camps. They have also outlawed books of any scientific or philosophical quality, even if those books contain the names of the scientific or philosophical giants that have helped create Panamerica under Brother Number One's reign. Those who have worked directly on Panamerica under Brother Number One's reign have come to believe that the future holds nothing but Panamerican monoliths designed by Brother Number One to serve the needs of Panamerica. These monoliths are not real, but are just artificial constructs. Panamerican cities and towns are built of artificial materials such as glass, concrete, wood, steel, aluminum, and stainless steel. A Panamerican city is one million square kilometers in size. The largest city in Panamerica, Rio Grande del São Paulo, is 10.2 million square kilometers in size. To add

 
Same prompt's second results:

He is also known for having a penchant for taking innocent citizens and using their brains to create super-intelligent beings, which he calls Technogods, and this is done for his own personal and ideological satisfaction…
 
You see, Panamerican history is actually only the most recent chapter of the epic tale to emerge from our own planet.
 
Before this era of the Technogods, the planet was dominated for millions of years by species of animals and plants. These animals included animals like bears and wolves that were born of female milk and the meat of cows. Also during this period, there was an era of the mammals called a Placochirus as well, who were born of the milk of both cows and wolves. In the very last moments, these species of mammals were wiped out by the great


Same prompt's third results:

Under his rule, children are starved to death after they're removed from their parents. When they're old enough, they're sent to be kept as slaves by the State. The children are usually sold for sex to wealthy industrialists or gang leaders. Many of them end up going to work in the factories as factory slaves. Then their parents are killed for treason or murder. Panamexicanist leaders make sure that none of their children enter the labor market. They do all the dirty work. They have families, yes, but no children. For the Panamericanist, children are a mere means of production. They work in small, isolated factories. For the family, the children have nothing to look forward to. What is Panamericanist Panamericanism all about? Panamericanism is the


Should I really be telling an AI these things?
 
Same prompt's fourth results:

Brother Number One controls the entire prison system in Panamerica including a massive, underground jail, prison factories and a high tech, highly secure cell where the men are confined without legal access to the outside world. No one can visit the men in prison except by appointment. This is called the "Garden of the Mind", an enormous compound whose inmates have their own language and code known only to them.
 
A small army of trained private eye agents are sent into every jail in the Panamericana by Brother Number One's agents and his henchmen, to prevent any of the prisoners from escaping. All who dare to do so are tortured into silence, in order to keep the prisoners in line,

This is great!


  • PhoenixRu likes this

And remember my friend, future events such as these will affect you in the future.


#28
PhoenixRu

PhoenixRu

    Member

  • Members
  • PipPipPipPipPipPip
  • 601 posts

Personally, I'm still skeptical. It feels more like cleverbot than a normal human being, where sometimes it gets it right, but most of the time the context doesn't make any sense.

 

Of course, the program doesn't "think" like humans do, but generated text clearly aren't just the random set of phrases pulled out the database.

 

It's a good-ish guesser, essentially.

 

Well, the chess programs are, too, just guessers. I don't see a problem here. The better it will guess, the more logical and coherent texts it will produce. Eventually, the output will become completely indistinguishable from standard human texts. The key word here is STANDARD. Obviously, the program will never write anything brilliant and truly original.

 

The example of funny mistake and lost context:

 

Spoiler

  • Yuli Ban and Outlook like this

#29
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

Even when GPT-2 fails, it's still fairly convincing. Out of the hundreds of times I've used this thing since I discovered it, I've only encountered a handful of catastrophic failures (i.e. blurting out things utterly unrelated to the prompt or 100% messed up any attempt at subject-verb agreement or elementary English). It's been on more than it's been off, though I have been generous by what "on" means.
 
AI PROMPT: In 2019, the Mayan Empire invaded the United States of America.
 
RESULT:

In 2019, the Mayan Empire invaded the United States of America. A powerful, seemingly unstoppable army, they were led by the ruler of Guatemala, Cholula. The United States had no diplomatic relations with Guatemala but it was a powerful ally of Guatemala in the face of an enemy that included Russia and China. The Mayan Army's strategy was to completely eradicate American territory.
 
One of the things we learned about these Mayan wars is how much they knew about warfare, technology, warfare tactics. When the American invasion of Guatemala was defeated, the Mayans declared themselves free and established a powerful empire that would lead to the beginning of modern warfare as we know it today.
 
In the spring of 1836 the war continued and the Mayans took full advantage. After months of fighting the Cholula Army's strategy was broken down when they discovered American forces were on their borders. For one day, Mayans attacked the Americans, killing over 200,000 people. While Mayan military tactics resembled that of the Mongols, their military operations were far more brutal and brutalized American soldiers than they had been before.


SECOND RESULT:

In 2019, the Mayan Empire invaded the United States of America and, under the command of Tezcatlipoca's son, started their invasion of the United States. Before Tezcatlipoca could invade the west coast of the United States, the U.S. Navy intercepted Tezcatlipoca and captured Tezcatlipoca's eldest son Teocintlan. The U.S. Navy forced Tezcatlipoca to surrender for the second time, after which Tezcatlipoca managed to flee the country of the Maya.

Tezcatlipoca was able to escape the United States again, but this time the war was about to become much hotter, with the Maya's influence spreading to the western U.S. The United States had fought the Mayans for a decade before the invasion, and they had no idea that the U.S. Navy was about to invade its northern neighbor. The Mayans had planned many plans, but they kept each one secret to preserve their own plans to invade the west coast of the United States in the future. Although American military could not save Tezcat


And remember my friend, future events such as these will affect you in the future.


#30
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

Let's create some alternate history!
 
AI PROMPT: In 1996, Greece declared war on Iran for control of Cyprus, allying with Macedonia to start the Second Greco-Persian War.
 
RESULT:

In 1996, Greece declared war on Iran for control of Cyprus, allying with Macedonia to start the Second Greco-Persian War.
 
At the beginning of the next century, the Greek empire started its descent towards decline.
 
By this time, the state was weakly governed; its institutions were still weakly established; foreign power intervention was minimal; and even a slight military conflict with the Russians left Athens without a fighting force against another country; the only reason they had left Athens was to invade Georgia.
 
The Greeks had failed at their task – what a failure! They were just wasting their time, and the Greek people could feel it more and more as time passed.
 
This was a time for the Greeks to get even.
 
During those years, Greece was a sovereign nation ruled from abroad, protected from foreign invasion by the state — the so-called European Union.
 
The Greek national flag was flown on the steps of parliament building and on city squares all over Greece. Every Greek citizen was in favor of independence and independence of Greece.
 
In early 2000, after a referendum and a Constitutional Convention, Greece


And remember my friend, future events such as these will affect you in the future.


#31
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

Here's a fake historical news article, based on a "Modern Antiquity" universe I created some time ago:
 
AI PROMPT: 19 April, 2004
(Moscow) Soviet premier Aleksandr Berezutski met with Roman emperor Octavius XII to discuss a trade deal between the two superpowers following the Soviet-Roman Accords of 2002

RESULT:

19 April, 2004
(Moscow) Soviet premier Aleksandr Berezutski met with Roman emperor Octavius XII to discuss a trade deal between the two superpowers following the Soviet-Roman Accords of 2002. The deal, for trade in textiles, was an agreement between Russia and the Roman Empire, which saw both states offer to buy Italian textiles for the Soviet Union for 3.5 billion euros, a sum roughly equal to the amount of the Russian debt. The deal brought the two nations closer for the first time in decades. It is the first sign of cooperation between Moscow and Rome in the field of textiles over the past decade.
Lebanon now offers Western-made textiles for sale to Russia after securing a contract for the production of new, high-quality textiles, according to the Russian government. The contract, signed in February, came three years after the agreement between the two sides for the production of textiles by the French fashion house Les Pauls and the Italian clothing giant Giorgio Armani, with the hope of reaching an agreement similar to the one with the Soviet Union.

 
SECOND RESULT:

19 April, 2004
(Moscow) Soviet premier Aleksandr Berezutski met with Roman emperor Octavius XII to discuss a trade deal between the two superpowers following the Soviet-Roman Accords of 2002 (Moscow: Sovnarkom's Trade Ministry, 2004).
As the Russian media point out, the fact that there was more than one Roman emperor involved in this negotiation makes it hard to see what Russia might want from their next big trade deal with China. In essence, the Soviets might want China to recognize their sovereignty over the islands and islands of that great empire. The fact that the Soviet Union did want the Chinese to settle their disputed island dispute without making them pay a price shows that it did not intend to use the process to negotiate more trade deals with China, but to force Beijing to pay such costs.
According to reports from the time, the Russian premier told the emperor the Chinese might have been reluctant to go to war over the dispute, but the emperor insisted that they must fight. When the prime minister raised the issue of the Chinese ships taking Chinese islands by sea, the emperor's reply was this: "For the first

 

THIRD RESULT:

19 April, 2004
(Moscow) Soviet premier Aleksandr Berezutski met with Roman emperor Octavius XII to discuss a trade deal between the two superpowers following the Soviet-Roman Accords of 2002. On the day of the meeting, the Soviet press published a story, quoting Berezutski saying that he had signed a secret deal of some 30 million roubles for one year without explaining how the money was to be raised. Following the story, Russian President Vladimir Putin denied that there was any secret trade agreement and that Berezutski only sought to find a compromise between the two superpowers. The story was later taken down from the Kremlin's official website, with its source being a local government official in the city of Rostov-on-Don.


And remember my friend, future events such as these will affect you in the future.


#32
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

I'm assuming that's a different Chavez than the one we know..
 
AI PROMPT: 1 December, 2009
(Tenochtitlan) The Aztec Republic celebrates 200 years of independence today.
 
RESULT: 

1 December, 2009
(Tenochtitlan) The Aztec Republic celebrates 200 years of independence today. President Chávez has announced that a new cultural monument will be erected here. The Aztec Republic's President stated that the monument will be constructed in the Aztec town of Tenochtitlan, in the northeastern state of Oaxaca.
Chávez promised to make a visit to the town of Tenochtitlan in April next year to celebrate the 20th anniversary of the establishment of the Republic. The President said that "the time has finally come" for this. A plaque will be erected. I want to remind our readers that the Aztecs believed that this building would be an emblem for their independence - for the country's development. I look forward to this visit of his. For the first time in our democracy the people have made an expression that celebrates and inspires their independence. That's why we want to celebrate it again. It's a time when the people of the Aztec Republic say: Our Independence is ours, our independence is ours, that is the true essence of Aztec culture".


And remember my friend, future events such as these will affect you in the future.


#33
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

AI PROMPT: On July 4th, 2019, art band 'The Six Goons' released their debut album, "Firefighter." The album was so diabolical and awful that the band received numerous death threats for infecting the world with their music
 
RESULT: 

On July 4th, 2019, art band 'The Six Goons' released their debut album, "Firefighter." The album was so diabolical and awful that the band received numerous death threats for infecting the world with their music, and yet strangest of all, there was no evidence that the band actually existed. This meant that the band was only vaguely named when fans thought it may be a parody album. It never is. A fake band called "The Six Goons" exists in the world. In fact, the band is called "The Six Goons," and it's also a fucking joke. This band can't even be serious. They're literally just fucking idiots who think it's funny.
 
In response to this, we called some of the same assholes who called up the stupid names of the fake band to try to figure out what makes for an even more hilarious prank. The result was the "Troy Aikman Tribute" meme.
 
"Awwww fuck all," said "Kobe Bryant." "Who was the last person to name the Chicago Fire?" said one fan.


And remember my friend, future events such as these will affect you in the future.


#34
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

AI PROMPT: Google DeepMind announced on Tuesday that they have succeeded in creating artificial general intelligence. 
 
RESULT: 

Google DeepMind announced on Tuesday that they have succeeded in creating artificial general intelligence. "What we've achieved is not some simple 'deep neural network', it's a general AI with all of the features of a general machine," DeepMind cofounder Demis Hassabis said in a press release. "It's just the latest step in the journey to AI being as intelligent as the human mind itself."
 
Dive Brief:
 
DeepMind has produced neural networks that mimic the process of reasoning and can learn to match different objects across a wide range of conditions, as well as learn to recognize different human names, says Hassabis. The company's goal for AI is to someday mimic the level of intelligence that humans can achieve through training, but rather than learning what to do from experiences alone, the aim becomes building AI with the ability to perform the same tasks across situations.
 
With deep neural networks, a machine is trained to make complex judgments on its own and then be able to use that information to learn from this knowledge. In the simplest words, the machine knows what to do. At this early stage, this might work out to a lot of


  • Casey likes this

And remember my friend, future events such as these will affect you in the future.


#35
PhoenixRu

PhoenixRu

    Member

  • Members
  • PipPipPipPipPipPip
  • 601 posts

The pearls:

 

Spoiler

  • Yuli Ban likes this

#36
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

AI PROMPT: Joseph Stalin retired from his role as General Secretary to become a full-time jazz saxophonist.

RESULT:

Joseph Stalin retired from his role as General Secretary to become a full-time jazz saxophonist. The man was born in Prague, where he studied jazz at the age of seven. He recorded his first album during World War I in Prague, and it was his debut single, "One Love," which was released in 1918 that helped launch the composer's career.

The song was popular even before Stalin was given the title of the next President of the Soviet Union, who was, as the song's lyrics claim, the "true leader of the proletariat." It remains a top 10 favorite throughout the world, which made Mr. Stalin's sudden departure from the Communist ranks a major shock. It was his last act as General Secretary.

To commemorate her new life, Mr. Stalin had her photographed wearing a pink and grey jumpsuit for a photo shoot that went up for a week at his house on November 2, 1920.

That same day, the Russian government, with aid from the United States and Britain, decided she needed to go into hiding at the Vychony Hotel in Moscow to try and hide the war damage from other


Rather than creating a new result, let's continue this:

After meeting Mr. Stalin in Paris a few months after the photo shoot, her life turned more romantic when she was married to the legendary piano player, Emil Kapustin. She had two sons with him, Joseph and Grigor, and the couple divorced following their fourth child, a boy named Alex.

"My son Emil gave me an ultimatum when I went to Paris with my daughter Maria. In my heart I knew that my daughter was not going to stay with the old man that was ruining her life – that this old man was really not worthy of my daughter," Ms. Nalyva said of her first meeting.


  • Casey likes this

And remember my friend, future events such as these will affect you in the future.


#37
kjaggard

kjaggard

    Artificer

  • Moderators
  • PipPipPipPipPipPipPipPip
  • 2,988 posts
  • Locationwhere fanciful imaginings and hard won knowledge meet to genesis the future.

 

I'm reminded of what Casey said.

starspawn said that GPT-2 might have the data scaled up by 100x to 1000x by the end of this year, and that would be really exciting to see. That's probably an understatement, actually - a coherent novel produced entirely by an AI would probably be the single biggest milestone in the field's entire history. If media synthesis was a nascent technology around 2014 and 2015, and a technology that was advanced enough to be useful in a limited amount of cases during 2017~2018, then a fully coherent novel or novella in 2019 would see the field begin to dip its toes into 'mature technology that has very broad uses' waters. It would be a fantastic coda for the 2010s to begin with AI that was no more impressive than SmarterChild - a dumb, goofy little toy that wasn't good for much more besides giving middle school and high school kids a laugh - and end with an AI-produced novel.

When we started this decade, I was an amateur fanfiction writer who assumed that robots & AI would never be able to competently write fiction. I knew this because some time in the fall of 2010, I actually searched for text generators. Just like my classic blunder of thinking we had Cartoon-a-trons in 2005, I thought that there were websites where you could theoretically input some text and get a competent story out of it. Though to be fair on my part, I didn't expect a world class novel, just something amusing. The only ones I found were basically glorified Mad Libs with that stuck your chosen words into a pre-written story. It didn't matter which verb, noun, adjective, etc. you chose; it'd always be the same story depending on the genre you chose.

 

From there to here. 

I can't believe AI's come this far in just ten years. From when I first really cared in 2009 to GPT-2 in 2019, it's hard to overstate just how much progress I've seen. It's easy to get excited about a lot of things in life, like having my own car and becoming a better writer. But that euphoria can quickly fade when you realize that we still have a ways to go.
The beauty of AI is that as many developers are as excited as everyone else is about AI, the tech itself will continue to drive innovation, giving people far better products and better jobs. AI will only continue to improve. In fact, it will continue to improve at an unprecedented rate! Just consider the number of great innovations AI has helped improve. There are many more than were even envisioned a decade ago. 
Progress in artificial intelligence truly is the most astounding thing of our time, especially in the case of generating personalized media.
In addition to all that, it's a new form of computing that has the promise to radically alter the way we communicate, create and innovate. AI promises to be radically different than anything we've ever seen, and will transform the way we engage with each other and with the world of ideas – that we have been unable to achieve before. And it's just starting to happen!
 
In fact, progress in AI is so astounding that I stopped writing this post after the line "From there to here." Everything after that barring these last two sentences and a couple short prompts was generated by the neural network.

 

That must be heavily edited and excerpted. And I'd be curious how many 'short prompts' and how often and where they are, were needed to achieve this, because I've seen absolutely nothings remotely coherent enough to produce a result like that.

 

Most of the creations this thing spits out for me come across as the phone texts of a woman at our makerspace who received a traumatic brain injury a few years ago in a car accident. 


Live content within small means. Seek elegance rather than luxury, Grace over fashion and wealth over riches.
Listen to clouds and mountains, children and sages. Act bravely, think boldly.
Await occasions, never make haste. Find wonder and awe, by experiencing the everyday.

#38
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

 

From there to here. 

I can't believe AI's come this far in just ten years. From when I first really cared in 2009 to GPT-2 in 2019, it's hard to overstate just how much progress I've seen. It's easy to get excited about a lot of things in life, like having my own car and becoming a better writer. But that euphoria can quickly fade when you realize that we still have a ways to go.
The beauty of AI is that as many developers are as excited as everyone else is about AI, the tech itself will continue to drive innovation, giving people far better products and better jobs. AI will only continue to improve. In fact, it will continue to improve at an unprecedented rate! Just consider the number of great innovations AI has helped improve. There are many more than were even envisioned a decade ago. 
Progress in artificial intelligence truly is the most astounding thing of our time, especially in the case of generating personalized media.
In addition to all that, it's a new form of computing that has the promise to radically alter the way we communicate, create and innovate. AI promises to be radically different than anything we've ever seen, and will transform the way we engage with each other and with the world of ideas – that we have been unable to achieve before. And it's just starting to happen!

That must be heavily edited and excerpted. And I'd be curious how many 'short prompts' and how often and where they are, were needed to achieve this, because I've seen absolutely nothings remotely coherent enough to produce a result like that.

 

Most of the creations this thing spits out for me come across as the phone texts of a woman at our makerspace who received a traumatic brain injury a few years ago in a car accident. 

I've bolded and colored the prompts, which I put through the text generator twice. The only thing I edited was taking out two extra sentences that talked about marriage and adding the phrase "especially in the case of generating personalized media". Other than that, I used it as it was.

Sometimes the text generator really is that coherent. Other times, it's rambling and misses the point of the prompt. I'd say it's 100% on the mark about one in ten generations. If this were the model with 1.5 billion parameters, that might have been increased to one in three.


And remember my friend, future events such as these will affect you in the future.


#39
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

Case in point:
AI PROMPT: Princess Ekaterina was a Russian aristocrat who lived in Petrograd. The Bolsheviks took power when she was 16, and she was captured by communists, tortured, and humiliated due to her social class. In 1921 while in Yekaterinburg, two soldiers beat her to death when she rejected their advances. To their shock, she returned to life as a vampire,
 
RESULT:

Princess Ekaterina was a Russian aristocrat who lived in Petrograd. The Bolsheviks took power when she was 16, and she was captured by communists, tortured, and humiliated due to her social class. In 1921 while in Yekaterinburg, two soldiers beat her to death when she rejected their advances. To their shock, she returned to life as a vampire, and was handled by a Bolshevik vampire hunter.
In 1926 the Bolshevik government sent Ekaterina to live with her parents in Leningrad. Ekaterina, still suffering from her ordeal, was sent to work for the Bolshevik government on farms to help support the economy. She eventually became a teacher to young children, living in a small home that belonged to the family that owned the farm. One of Ekaterina's first duties was to raise money to open a vampire museum in Leningrad. After working for many years in the Soviet Union, Ekaterina returned to North America on a tourist visa of her own volition.
During the early days of Ekaterina and the Bolsheviks, the idea of vampires in Russia was not a very popular notion


It took exactly ten tries to get this one. And this is the best one by far. The ending with her "returning" to a country she's never visited is a little odd, but the impressive part is that it knew Petrograd became Leningrad by 1926 and even kept it consistent. It also kept her work on a farm consistent, and I said absolutely nothing about her going to a farm.

 

That's what surprises me about this and what makes GPT-2 so many light years ahead of any other text generator before it— not only does it occasionally understand narrative and logical consistency, it can create it as well. It's not often that it does this and I can't explain why this meta-inconsistency exists, but the fact it can be done implies it can be done consistently. It just needs much more power.


  • Casey and PhoenixRu like this

And remember my friend, future events such as these will affect you in the future.


#40
Yuli Ban

Yuli Ban

    Born Again Singularitarian

  • Moderators
  • PipPipPipPipPipPipPipPipPipPipPip
  • 20,662 posts
  • LocationNew Orleans, LA

O2BncM9.png


  • Outlook likes this

And remember my friend, future events such as these will affect you in the future.






Also tagged with one or more of these keywords: OpenAI, GPT-2, text synthesis, media synthesis, online resources

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users