February 16, 2019

"Feed it the opening line of George Orwell’s Nineteen Eighty-Four – 'It was a bright cold day in April, and the clocks were striking thirteen'..."

"... and the system recognises the vaguely futuristic tone and the novelistic style, and continues with: 'I was in my car on my way to a new job in Seattle. I put the gas in, put the key in, and then I let it run. I just imagined what the day would be like. A hundred years from now. In 2045, I was a teacher in some school in a poor part of rural China. I started with Chinese history and history of science."

From "New AI fake text generator may be too dangerous to release, say creators/The Elon Musk-backed nonprofit company OpenAI declines to release research publicly for fear of misuse" (The Guardian).

20 comments:

Freeman Hunt said...

News is already full of fake news. Journalists did that themselves without the help of AI.

Mark said...

Of course, Orwell's Britain uses a 24-hour clock, so that's hardly futuristic (although it did sound strange when I read it back in the 70s).

Jeff said...

So the algorithm can write sentences, but when you string the sentences together you get a paragraph that looks like it's author had a bad case of ADHD. So what?

Churchy LaFemme: said...

It's no Mentifex!

John henry said...

I've seen this and wondered what is the danger?

Why is it too dangerous to release? Because it won't live up to the hype and embarrass Elon?

I call bullshit. Not on the generator. Those exist. On the dangerousness.

John Henry

rhhardin said...

Hype. It's a markov system. What in the database matches what we just said most closely, output the next bit of that. Repeat with the new output state.

Refinements would be mostly using something other than words as they happen but something parsed, as the current state.

Leland said...

For the first time, I'm impressed with AI.

Fernandinande said...

"ThisPersonDoesNotExist.com uses AI to generate endless fake faces" (reload ThisPersonDoesNotExist.com to get a new face).

Chris Lopes said...

The thing is, the AI has no understanding of what it wrote. Computers are really good at simulating intelligence (much like our elected officials), but it is simply a trick. For perspective, I advise googling Eliza.

rhhardin said...

...desire to "join in." Everybody has, at one time or another, experienced the urge to throw a plate of jelly or a half grapefruit, an urge comparable to the inclination that suddenly assails one to leap from high places. Usually this tendency passes as quickly as it comes, but it is astounding how rapidly it can be converted into action once the spell of dignity and well-bred reserve is broken by the sight of, say, a green-glass salad plate flying through the air. It is all but impossible to sit quietly by while someone is throwing salad plates. One is stirred to participation not only by the swift progress of the objects and their crash as they hit something, but also by the cries of "Whammy!" and Whoop!" with which most men accompany the act of hurling plates. In the end someone is bound to be caught over the eye by a badly aimed plate and rendered unconscious.

My contemporary, Mr. Fowler, in a painstaking analysis of the split infinitive, divides the English-speaking world into five classes as regards this construction: those who don't know and don't care, those who don't know and do care, those who know and approve, those who know and condemn, and those who know and discriminate. (The fact that there was no transition at all between the preceding paragraph and this one does not mean that I did not try, in several different ways, to get back to the split infinitive logically. As in a bridge hand, the absence of a re-entry is not always the fault of the man who is playing the hand, but of the way the cards lie in the dummy. To say more would only make it more difficult than it now is, if possible, to get back to Mr. Fowler.) Mr. Fowler's point is, of course, that there are good split infinitives and bad ones. For instance, he contends that it

http://downwithtyranny.blogspot.com/2009/12/thurber-tonight-ladies-and-gentlemens_23.html

Thurber faking an awkward transition.

gilbar said...

Chris Lopes said...

The thing is, the AI has no understanding of what it wrote. Computers are really good at simulating intelligence (much like our elected officials), but it is simply a trick. For perspective, I advise googling Eliza.

* For perspective, I advise googling Eliza.
> Don't you ever say Hello?
* For perspective, I advise googling Eliza.
> Come, come, elucidate your thoughts.
* For perspective, I advise googling Eliza.
> Do you expect a different answer by repeating yourself?
* For perspective, I advise googling Eliza.
> Why did you repeat yourself?

Sebastian said...

Turing got it wrong.

Some progs hear the output of "Kathy Griffin" and think she's human.

Chris Lopes said...

@gilbar
Exactly. LOL

Yancey Ward said...

It was the best of times, it was the worst of the times. James had finally convinced "Halloween Sashimi Slicer 1970" from Match.com to agree to a in-person date. Little did he realize that Lorena Bobbitt had found her next victim.

tim maguire said...

Elon Musk's fake text generator thinks that in the future cars will run on gas and use keys to start?!

Does Musk know about this?

roger said...

Call me Ishmael..........

Henry said...

Pawn to King 4.

FullMoon said...


Blogger Henry said...

Pawn to King 4.


Worked for Bobby

Oso Negro said...

On the bright side, news publishers can save a fortune on journalist salaries. And I’m asked to prove that I’m not a robot

stlcdr said...

So, what exactly is the fear?

I suspect that it will generate a story which doesn’t fit the current narrative.