I fed a number of interviews with Donald Trump into a recurrent neural network. An RNN can learn to predict what letter ought to come next, depending on how well it has learned. The more data you can feed it, the more working memory your machine has, the better the eventual results. Once you’ve trained the model, you sample from the model and see what kind of text gets generated. If you turn the ‘temperature’ way down low – ie, you don’t let it get creative in its guesses – you end up with a kind of platonic ideal of what the training texts were about. The computer doesn’t really ‘know’ what the text is about; just, given whatever letter it starts with, the next likely letter. You can ‘prime’ the results by giving it a short bit of text to start the sampling from. So let’s see what my wee machine learned. In the dialogue below, my questions are the primetexts I used. The temperature is set to .5, meaning the machine is evenly balanced between fidelity & creativity in its responses. Since I copied interview texts verbatim, there are also other interlocutors in the text below, digital-doppleganger-reporters. Nothing of what follows involved any other humans except me; the texts are copied verbatim from the output of the machine, which learned by ingesting the first couple of pages of results of a search for ‘trump interview transcript’. Please don’t make me search for anything Trump related ever again.
SG: How will you govern?
COOPER: They have to do it.
TRUMP: Maybe they say this.
COOPER: You see the end of the worst and then I never have to be president of the world in the something with the first — we are going to have to have to be on the history of the people where you see the world but the experience that you’re not going to be able to have the polls that was a real problem with the polls and they are not going to do it.
TRUMP: I know.
MATTHEWS: Well, you don’t see that.
TRUMP: I mean, I think it’s totally cared what I was a lot of audience. I mean, you know, when you want to talk about the party, and they don’t want to say that. Because we have to do it. I have to do it. I have to be able to be a majority and they have to be able to do it anymore. He was a debtor nation, and we’re pro-life. I pay them in a politician, when I was a pretty much different than the banks. And you know, I did believe in the country is a lot of excellent against a great state to Japan. It’s not a strong with the states and they don’t like that.
MATTHEWS: Well, the party of the problems that you have to still the formibility of people that was the some that I say it was a problem with the world or the same people that you’re going to be able to be a very much about when you say that?
TRUMP: I think they’re going to have a strong question.
MATTHEWS: Well, you have to look at the establishment that are the problem.
MATTHEWS: I know.
TRUMP: Yes, I don’t want to say this. They should protect them on which people with the money to meeting with the Middle East. They want to like to get it to the world. What do you actually have to get a lot of people in the problems we have a strong and we’re going to have a strong convention and we’re going to be a problem with the terror. I have to be able to do it. And I would go to the federal government in the media, when they don’t say that. I can do that. And I know you have to stop the Japanese.
SG: Why do you have to stop the Japanese?
TRUMP: I was a lot of people are looking at the polls who was a deal. We’re going to see the law in the hard of the states, they’re excellent. And I don’t think they’re a good president. I think I think that they put the banks, because I have all over the country. I don’t know. And I think it’s a great place to be a book about it. And they can’t believe it because we have to say that they’re going to be crazy because they have to do it. I have a fact for the guy that they can’t do that.
TRUMP: They’re not doing something we have to stop what he’s wonderful, I think it’s a problem with the party of the states. That’s not going to have a lot of people are running for the problem with the polls that I don’t know what at the things about them in the United States and then they don’t want to take it pretty world. I have the place that they’re a very good deals. We can be the only one that are going to say that.
MATTHEWS: I think you do think.
TRUMP: You know, they’re going to be a president of the things. And I don’t want to support me. I think they’re going to have a second interview that we have a strong people, and he was a strong argument that they say, “I don’t know. I was a very rich country. We have to be able to be probably from the way they say, just the problem is a strong question. I am everybody else, and I was against the Republican Party and we’re going to look at the world, and they were in the same pollsters and they have to stop the worst institution. You know, we want to be a strong again.
MATTHEWS: I know the country, the morning is going to be able to win. But it’s not a first thing to the party of the United States, and the presidential party and the ground.
TRUMP: I think it’s not going to be the case. I was a strong problem that they don’t know what happened.
SG: What about the Canadian Border?
TRUMP: I think I think that they really want to have to be able to really started up and we’re better off that country the new one of the reason that we have to do is the banks and saying, “Why do you take a look at her way. I think they don’t want to have a tremendous party of the United States. I don’t know what they have to start to him and really say you have no days and they don’t know what I would be in the same time when you have a lot of people saw the problems where they get away with the end of the worst and I don’t want to do it.
COOPER: But they have a country has on that person. I have groups of people in the health care of the most angressive to the federal government in the Republicans would be a big thing that the last debate.
TRUMP: I have a poll that they’re going to have to do it.
COOPER: I would say, you say, “That’s what I think that you’re going to have to make a tape that it’s a strong question in the case of the world. They were the state and the view of the polls that you have to be able to say this?
MATTHEWS: I would have said you said, “You said — who was a great global warming, there’s a lot of people, but I would take a look at the problem.
TRUMP: You can go to the polls that I am something that doesn’t have to do it. John, what is this point.
MATTHEWS: OK. And I was about the way the problems that you’re going to have to be able to not be very well.
TRUMP: Well, I would be very strong for the problems what I don’t want to say that it’s in an antwere because they don’t want to be the case. I was a strong problem that they don’t know what happened.
SG: Most of this makes no sense, eh?
TRUMP: I have a lot of money.
I dialled the temperature down to .20, to see what the platonic ideal Trump might say, with no primetext this time…
COOPER: I want to say this is a strong argument that you would say that they want to do it.
TRUMP: I have to be able to be a lot of people are supposed to be a lot of people are saying that they’re going to be able to do it. I don’t want to say that. I was a great problem. I don’t know what he was a very strong and they get away with the problems that I was a strong angle to do it.
COOPER: But you said you don’t want to have to be able to do it.
TRUMP: I think I’m not going to be able to walk. I think I would say this is a great person. I mean, I would be able to say that they’re going to be able to be a lot of people are going to be able to do it. I was a president who was a president that they don’t know what he was a very strong with the people in the states and they don’t want to say this because they want to say that, when you say that they’re going to be able to be a debtor nation. I don’t want to be a lot of people are going to be able to be a lot of people who want to do it.
MATTHEWS: Why do you think that they say that they don’t want to say that.
TRUMP: I think they’re going to be able to be a lot of people are going to be a very strong and they’re going to be able to be a lot of people that have to stop the problems in the world is the way they have to be able to do it.
MATTHEWS: I know.
TRUMP: I don’t know.
MATTHEWS: What do you think it’s a lot of people in the problem with the problems that the polls who was a very strong and they don’t want to have to do it anymore.
TRUMP: You know, what do you say to the country. And I think they don’t want to do it.
TRUMP: I don’t know what they do that.
MATTHEWS: Well, you have to be able to say this is a debtor nation.
TRUMP: I don’t want to have to be able to do it because they don’t want to say that. I have to be able to do it. I mean I don’t want to say that.
Finally, I dialled it all the way up to .8, to let the Recurrent Neural Networked Trump loose…
And if he grabbed me, I have the numbers are going to look at the position for decides…
TRUMP: I can do that. I think because I don’t want to do that contract with a great will come out on my add.
MATTHEWS: Well, we’re pro-life and some of the views are doing that. And I may be here it. She’s a great sense to me what this is a real estate political problem.
TRUMP: Well, a couple of things here ago, you saw the question. You can grabbed you something you’re not pretty well.
MATTHEWS: No, but you have to do it. They talk about the numbers and the other party people will be losing a lot of family in the Republican nomination,
TRUMP: What did they’re going to be generals. And I was against our military — what does he said it’s just an example, he was occusions in the wall. I think this is Your fantastic. We’re horrible thing to stop abortion in shirt of the world. We have to do is all these states are running. But we shouldn’t have show a couple of months of issues that to do it. I mean, a lot of people don’t know what we’re seeing it, I made the problems in this country
It’s worth re-iterating that the quality of the output depends on the input and on the size and complexity of the calculated neural network. Your mileage may vary. Regardless: What’s the point of doing all of this? There’s an element of the Dada to it, an element of the performative nature of Oulipo type texts and ‘deformed digital humanities’. It makes something new, and in the juxtapositions and ludicrous ravings of the machine, I think we get deeper glimpses. After all, a recurrent neural network is close kin to so many of the other machine learning algorithms out there that (with the power and resources of major corporations or governments behind them) are starting to imprison our lives. Do you want this sort of thing deciding who gets the loan, who doesn’t? Who gets targeted for observation, and who doesn’t? Here, the results are risible: but scale upwards…