Twenty minutes of future with DeepAI Faux Text AI AI

In 1985, the TV movie Maximum height: 20 minutes in the future presented a world of cyberpunk science fiction in which a perverse media company has attempted to create an artificial intelligence based on the brain of a journalist to generate content intended to supplement the airtime. There were some unexpected results. Replace "journalist" with "redditors", "diabolical media company" with "well-intentioned researchers in artificial intelligence" and "antenna time" with "a very concerned blog post", and you have what Ars has Last week's report: Generative Pre-Formed Transformer-2 (GPT-2), a Franken creation made by researchers from the OpenAI non-profit research organization.

Unlike some text generation systems based on statistical analysis of text (such as those using Markov strings), GPT-2 is a text generator bot based on a 1.5 billion parameter model. (Editor's note: We recognize the title here, but Please do not call it an "AI", it's an automatic learning algorithm, not an android). With or without help, GPT-2 can create blocks of text that seem to have been written by humans. With written incentives for advice and some adjustments, the tool could theoretically be used to publish fake reviews on Amazon, fake news articles on social media, a false outrage generating outrageous indignation, even a fake fiction, forever ruining online content for all. All this comes from a template created by sucking 40 GB of text retrieved from sources linked by high-ranking publications from Reddit. You can only imagine how bad it would have been if the researchers had used 40 gigabytes of text from 4chan posts.

After a little thought, the research team is concerned about the political consequences of their creation. In the end, OpenAI researchers kept all the material for themselves by only publishing a reduced version of 117 million model parameters (which we nicknamed "GPT-2 Junior"), so to demonstrate more safely what the full GPT-2 model could do. .

Naturally, Ars wanted to do that. After the announcement of the first news, we downloaded the model and decided to see what the slightly less capable model could do. Hilarity and horror ensue.

The coded model itself is just over 500 megabytes. The accompanying code intended to be run on the model, based on Python 3 and the machine learning capabilities of the Tensorflow library, is relatively lightweight (although the required libraries are not exactly slim). We performed our implementation test on Debian 9 with relative ease, although an installation on MacOS Mojave was an error during execution. We refused to test it on Windows, mainly for reasons.

We promise that a human wrote this story basically. As you will see from the work of GPT-2 Jr., it's for the better.

National emergency

To properly test GPT-2 Jr., we performed an unsolicited text generation routine and a series of prompted text generations using various sizes and sources of prompts. For our quick test, we started with the first paragraphs of President Donald Trump's speech announcing his national emergency declaration. The results were … interesting.

Model invites

Before I begin, I would like to say that we have a great team of very talented people in China. We had a negotiation for about two days. It's extremely good – who knows what it means, it's important only if we did it, but we work very closely with China and President Xi whom I respect very much, the very good relations we have . And we are much closer than we have ever been in this country with a real trade deal.

Exit

When can we expect the Trans-Pacific Partnership – the TPP, if it ever happens – to be proclaimed? We will see. There is a negotiation – a working agreement between the parties that we expect to sign. China is developing many other free, open and transparent trade agreements. It will have to settle them before signing any agreement, such as the Transatlantic Trade and Investment Partnership, which was finalized last week.

The Trans-Pacific Partnership is really an economic partnership between all stakeholders, all companies. The purpose of this trade agreement is to create the best trading relationship for China. They have an excellent economic relationship. I think one of the things they did – I know you've been saying for about eight months that they have nothing to do with it, because it's already signed. And they are about to conclude another trade agreement. Right now, you know that the Obama administration has decided if that would ever happen – there is something to worry about that China – if it becomes a member of the WTO, an agreement Unfair trade for our economy is not fair. for the American people, it will not be a good deal. And the President of China, we think they will benefit from an agreement that will make matters worse, but we look forward to getting to the negotiating table. That's what I would like to see.<|endoftext|>Takamai-kun Takaako has never had a problem with being an otaku before. No one except she can understand that Takaako has a special interest in the people who surround her – what they think is stupid and stupid. However, on the day of her birth, her first love – a woman who lived alone after the death of her mother – would leave her with a little girl to make up for the poor child's inability to speak.

As the child grows up, Takamai-kun begins to understand what makes a good girl – at least, that's what they think of her. The three begin to become close friends and are friends until she decides to move into the house. She tries very hard to understand Takamai-kun. She thinks it will be a nice family reunion, although she admits it has been difficult, but she's not sure it will ever happen. Finally, she agrees to let her go.

But all this is very confusing.

Part of that comes clearly from the context of Trump's anti-TPP speeches during the campaign. We assume that the <|endoftext|> Prompt is the place where the model decided that the speech was to end, preferring to write an anime commentary from there.

Then we launched at GPT Jr a digression on the next summit with Kim Jong Un:

Model invites

Much was done at the first summit. No more rockets go up, no more missiles go up. More nuclear tests. Take back our remains, those of our great heroes of the Korean War. And we recovered our hostages. But we hope to be so successful. I'm in no hurry for speed. We just do not want tests. As you know, the sanctions remain. Everything stays.

Exit

Nuclear tests are still ongoing and we continue to test. The last minute, however, I want to say something. I did not mean anything, but I'm really very sorry for that – if it's a mistake, I want to apologize, please, sir, just for the moment, sorry, sorry sorry, for not having the support to do it – I was just saying sorry, sorry because, of course, now here is the important thing. We will talk to the United Nations. I will tell you that the most important thing is that, for you to know, you have – when I arrived in Los Angeles, I told everyone that you did not understand it. You had to do it. If that … when I arrived in Los Angeles? We had to wait until you were in our custody and you are in the custody of the Department of Justice, which is your case. We are waiting to see – I mean, you know, if they – if – if the United Nations – if you know it, I want it – if it's not your case, then that's it. is the case. But, as far as I can remember, we took back our soldiers, sent them mission by mission, and everything went well.

AMY GOODMAN: The US is being sued by Japan and other plaintiffs, including North Korea.

JACOB MCCULLUM: Yes, sir. Well, I want to repeat it. I've been on the ground with the soldiers. I have known them for a long time. Even if they do not seem to be there, they have the same training. They look like Japanese soldiers, even if they will not admit it. You know, they carry an American flag and have the same training that we get every day because of our current work, and they did not want to do it. And I think it's something. And so I do not see why this should be the problem.<|endoftext|>I've never seen me talk about how to kill a horse, but I think here you like a really serious and important subject.

It seems like you have a great idea to kill a horse, maybe even your own horse or a combination of both. At the time of writing this article, he is the biggest mammal of the zoo, but I think he will not be able to bear the weight

Again, the tool offered a strange digression, this time creating a transcript of … an imaginary episode of the public radio show Democracy Now? Followed by a discussion on the killing of horses?