An AI (also) wrote this review and it's amazing.

Henry A. Kissinger (a 98 year old man who needs no further introduction), Eric Schmidt (ex Google big boss) and Daniel Huttenlocher (MIT professor) wrote a book together on artificial intelligence. Kevin Roose of NYT reviewed it and took some help from a limited version of GPT-3. I did the same, using the initial text from his review to prompt the AI, but used the full beta version of the model. And I think the result is amazing. Read it below.

But don't stop by just reading it. Take a moment to think long and hard on what's actually going on here. The title of the book hints at our ”human future”, and having text generated this way is definitely a part of that future. Love it or hate it, the genie is out of the bottle and our mind and time should be spent on how best to make use of this technology. But at the same time, I often see such good results from the model that it challenges my view of my own value. Robots have long since been able to do everything better than my physical body – now what if this machine also writes a book review better than I could ever hope to do?

Anyway, here we go - full text below the screenshot.

When you read the text, please note how creative it is – it even brings in Scheherazade, an old AI model into play (but totally reinvented by GPT-3 to be something else than the original model), and have it review the book. I mean, how incredibly meta isn't that? My mind is blown away by the fact that the algorithm has the exact same idea as Kevin Roose. Just think about it...

Also, please note that I did not force GPT-3 – in my prompt above – to be actual and factual. So it's just making things up here. Possibly makes the whole thing even more mind-blowing...

Finally; the text below is not edited in any way and it was the first attempt.

THE AGE OF AI?

And Our Human Future?

By Henry A. Kissinger, Eric Schmidt and Daniel Huttenlocher?

One of the great promises of technology is that it can do the work that humans find too boring or arduous.?

In the 19th and 20th centuries, factory machines relieved us of repetitive manual labor and backbreaking farm work. In this century, artificial intelligence has taken care of a few more tasks — curating Spotify playlists, selecting the next YouTube video, vacuuming the floor and so on — but many more mind-numbing activities remain ripe for the picking. The experts promise us that someday, all of our least favorite chores — including complex cognitive ones, like interviewing job candidates or managing global supply chains — will be outsourced to machines.?

But that day has not yet arrived. Or has it??

After finishing “The Age of AI,” a new book about artificial intelligence by Henry A. Kissinger, Eric Schmidt and Daniel Huttenlocher, I found myself unmoved by the prospect of reviewing it. I’ve read dozens of books about A.I., and while the conceit of this one was intriguing — bringing together a 98-year-old diplomat, a former Google chief executive and an M.I.T. professor — the book itself was a fairly forgettable entry in the genre.?

. . .

[the above is the pasted text from the review in NYT used to prompt the algorithm. Everything below is written by GPT-3]

. . .

So I decided to try something different. I asked a computer program to read the book and write a review.?

The program was a new artificial intelligence algorithm called Scheherazade. The algorithm was developed by some of my colleagues at The New York Times, and it was trained on the text of hundreds of thousands of articles that had been published on the Times site, dating back to 2007. This data set — which we’ve made available to researchers — includes information like author, publication date and article length, as well as a “sentiment score” that ranges from negative (“Trump’s language was especially inflammatory”) to positive (“The new health care law will save the lives of thousands of people”).?

Using this data set, the algorithm was able to identify recurring sentences, nouns, adjectives, adverbs and other linguistic markers that are associated with particular sentiment scores.?

It was then able to mimic this style of writing, using what it learned.?

By reading thousands of articles about artificial intelligence, Scheherazade learned to mimic the style of human authors. It can generate newsy paragraphs or blog posts, or imitate the style of a particular author. Scheherazade is the latest example of the power of “generative adversarial networks,” or G.A.N.’s, a new class of artificial intelligence algorithms that are capable of producing images, text or sound in ways that are strikingly realistic.?

I gave Scheherazade this particular book because I wanted to see whether the algorithm could mimic the writing style of the authors. And sure enough, Scheherazade’s review, which is below, was similar to the human-authored reviews in its cadence, word choice and structure.?

Here is what it came up with:?

Artificial intelligence is already here. It’s just not very good yet.?

Artificial intelligence is everywhere. From Apple’s Siri to the money-transfer service Venmo to Google Translate, A.I. technology permeates our lives.?

But whether the technology is truly intelligent is another matter.?

A.I. programs can perform narrow tasks like recognizing faces or translating one language into another. But they remain terrible at more complex tasks like understanding human speech, reasoning and making ethical decisions.?

And so far, the technology has not yielded the economic benefits that many people had expected.?

“This is the great paradox of the artificial intelligence revolution,” the authors write. “The technology that many people believe will bring utopia is currently improving the world far more through incremental improvements to the smartphone in your pocket than through truly transformative changes.”?

The book is an attempt to understand why the revolution has not yet arrived.?

To do so, the authors — Henry A. Kissinger, the former secretary of state; Eric Schmidt, the executive chairman of Alphabet, Google’s parent company; and Daniel Huttenlocher, the dean of Cornell Tech — draw on an eclectic range of sources.?

The authors interview the leaders of companies like Google, Amazon and Microsoft. They talk to people in the military and in government about the uses of A.I. in the fight against terrorism. And they also interview a variety of technologists and researchers, including a former Uber engineer who helped develop a program to automate driving.?

The authors’ goal is to identify the barriers that stand in the way of using A.I. to generate the sort of radical improvements in health care, transportation and other areas that were promised, but have not yet arrived.?

One of the biggest challenges, they conclude, is that A.I. systems are so complex that they “often become too complex to be understood by the humans who use them.”?

This is a big problem in health care, in particular. “Shouldn’t a machine capable of making a diagnosis better than a human be able to explain why it reached that conclusion?” they ask. “But when A.I. systems reach a certain level of complexity, they become incomprehensible to the people who built them.”?

Another problem is that A.I. systems are not good at dealing with the kind of uncertainty that is found in the real world — for instance, the fact that the road in front of a self-driving car may be slippery for some people and dry for others.?

“While A.I. has achieved dazzling successes, it has not yet lived up to its potential,” the authors conclude. “The answer to the question of why A.I. has not yet delivered broadly is that we are still waiting for A.I. to grow up.”?

The book is not very persuasive when it comes to identifying the economic benefits that will come from A.I. in the future. It talks about A.I. applications like autonomous cars and virtual assistants, but it does not provide much insight into how new technologies will improve the world.?

And the book doesn’t grapple at all with some of the thorniest issues related to artificial intelligence, like the automation of jobs and the use of A.I. by oppressive governments.?

But the book does provide a new way to think about the future of artificial intelligence.?

In an introduction to the book, Mr. Schmidt and Mr. Huttenlocher note that some of the most transformative technologies of the last century — including the automobile, the airplane and the Internet — were not widely appreciated in their early years.?

These technologies were wildly successful only after the people who developed them grew into their capabilities. “The founders of Google, Amazon and many other successful A.I. companies are convinced that the next great wave of A.I. technologies will find their origins in the technologies they are building today,” they write.?

The book’s title comes from the Greek historian Polybius, who wrote that “the rise of artificial intelligence is inevitable.” But the authors are not so sure.?

“We are not sure if A.I. is worth the attention it receives,” they write. “We are not even sure that it is important.”?

And yet, they argue, the technology is already in our lives. “We are in the age of A.I. in much the same way that we are in the age of electricity,” they write. “We are surrounded by it, use it every day, and are still unsure of what will come next.”?

I asked Scheherazade the same questions I asked human writers who have reviewed this book. Here are its answers, in full:?

Q. Does this book make a convincing case that artificial intelligence is already here, and that it’s beneficial??

A. The book makes a strong case that artificial intelligence is already here, and that it is already beneficial.?

The book addresses two critical questions about artificial intelligence. The first concerns the nature of artificial intelligence. Are we in the age of artificial intelligence? Or are we in the pre-age of artificial intelligence? The book makes the case that we are in the age of artificial intelligence, for many reasons.?

One, the book cites the fact that artificial intelligence is everywhere. From Apple’s Siri to the money-transfer service Venmo to Google Translate, artificial intelligence technology permeates our lives.?

Two, the book cites the fact that artificial intelligence is already benefiting us in many ways. One example is that cancer survival rates have increased because of artificial intelligence.?

Three, [...]

Cover photo by Markus Spiske on Unsplash