Guest Author Reinoud Kaasschieter
Computational Creativity is on the rise and everyday new examples are published. Computers help people create film trailers, musical compositions, and even write news articles. We should ask these three questions:
- However, can Artificial Intelligence (AI), as it is called, be truly creative by finding and publishing new content?
- What can we expect from computers when writing texts?
- Will creative computing eventually replace creators?
Natural Language Generation
Natural Language Generation (NLG) is taking off. NLG is part of the suite of Artificial Intelligence tools. NLG can create texts from bits of information, mostly listings with data. The lists primarily include information in a tabular format. For instance, sports results or share prices at the stock exchange largely come in tabular format. From those results, reports about the match or day at the exchange market are generated. You create a template for the kind of text you wish to create. Together with the listings it will create a story. A story that you can redact and finalize. However, in many cases, advanced texts merge, and all kinds of variable text elements are inserted in a template. Comparative challenges between a human text writer and one of those programs yielded rather convincing results, given how hard it was to distinguish between human- and machine-generated texts. Although the computer versions of texts lack the eloquence of the human versions, who needs that for sport results?
Should you welcome or fear the rise of intelligent machines? That depends on whether they will be working for you, or you will be working for them.
(Sarah O’Connor, Financial Times)
Will stories be original every time? Not really. However, we don’t always require originality when all we need are facts. Wordsmith for example, is actually used by big content firms. Hundreds of customers including Allstate, Microsoft, The Associated Press, and Yahoo! use Wordsmith to generate more than 1.5 billion pieces of content per year.
Do we really need (always) quality texts?
These applications don’t do quality in-depth journalism. Don’t expect investigative and explorative stories. This is not possible, because the computer only knows the data that you’ll feed it. It cannot go out and talk to people. But that’s not the aim of these services. The aim is to create the bulk of stories that you’ll find on those websites. Large corporations, like Coca-Cola, are currently experimenting to take this process further. Their experimentation will explore how AI can be applied to everything from choosing music to updating social media and even writing scripts, although removing humans from the equation entirely remains a ‘long-term vision’.
I don’t know if we can do it 100 percent with robots yet — maybe one day — but bots is the first expression of where that is going.
(Mariano Bosaz, Coca-Cola)
But let us remember that Artificial Intelligence is used for automating tasks, and since the dawn of the Industrial Revolution, mechanization and automation are used to take repetitive tasks away from humans. Creating content also involves quite a bit of repetition. When you examine your workday, how much time were you truly creative and attempting something new?
It goes without saying that you can also try to use computers to create fake news. Fake news is quite popular these days and seems lucrative. You can attract viewers with fake news and a click bait is set up easily. Will fraudsters use Artificial Intelligence to create phoney news? Of course they will. Criminals are always interested in the latest technology to stay innovative to generate new sources of revenue.
One thing is certain, fake news is threatening the degree to which people are informed about world events. (…) There is a role for AI to play in separating fact from fiction when it comes to news stories.
(Hira Saeed, VentureBeat)
In my view, a fake news arms race will emerge. Facebook is now trying to tackle fake news with AI. But the people in the Balkans writing those messages will also improve their skills to stay on top of their game. Artificial Intelligence isn’t good enough to detect fake news at the moment. But when it eventually does, the fake news generators will also use Artificial Intelligence. Predictably, a race between Artificial Intelligences lies not too far in the future.
Do fake news generators exist? Yes, in the academic world these generators have been built. But I’m afraid the current fake news producers are not that keen in sharing their knowledge.
Can computers create fiction?
Can computers generate original content? Artificial Intelligence is used in the film industry with some surprising results. The movies Sunspring and It’s No Game are interesting in this respect. Here, Artificial Intelligence has written the script. By analyzing patterns in existing science fiction movies, AI wrote its own new version. These are short science fiction movies, not quite the material for blockbusters. Interesting for nerds and science fiction geeks, but not for the general audience. But let me tell you, if you’re into avant-garde Dadaism, perhaps you’ll find the text of the script rather interesting. The text really doesn’t make sense, just garbled words. And I truly admire the actors who can act out those lines convincingly. Okay. These lines of text don’t make up film scripts that will earn instant box office successes. But we’re hopefully progressing.
Garbage in, garbage out
All the examples I have given you share one thing in common. You learn and train the computer in doing one thing. It is not a general purpose machine, not a replacement of a human with all its human capabilities, experiences, emotions and quirks. It only knows what it has been learned. It won’t investigate on its own, and doesn’t know synergy and lateral thinking. Some have attached such learning to the Internet as a real world knowledge base. Computers end up being racists which is a good result because racism on the Internet got picked up accurately by the computer.
So if you want a computer to create content, it cannot go beyond what is learned. When you try to do so, it will create utterly bogus results.
This is an excerpt of the presentation Reinoud Kaasschieter has given at the Information Energy 2017 conference on May 17th, 2017 in Utrecht, the Netherlands.
Cognitive computing in the arts
A more successful application of Artificial Intelligence is Cognitive Computing. Cognitive Computing can understand texts. Others say it’s just a pretty advanced search engine. We’re physically unable to read huge amounts of text. The computer can do this for us in seconds. IBM Watson can help artists, researchers and scientists find new insights into texts, stories, articles and so on. Things the reader didn’t know on beforehand. Insights that can inspire him or her to create works of art.
How does it work? You input a lot of text documents into the computer memory. You let the computer understand text by helping him analyze the semantics of the text: the words and their relations. Finally, the computer can use what he has learned to go through large amounts of other texts. Cognitive computers can also help content creators. It can help investigations searching for references in texts. It can provide answers to questions. But only in the field of knowledge that you have taught the computer. This is called narrow AI. Specializing in only one domain, on area of knowledge, but in great depth.
About the Author:
This guest article is written by Reinoud Kaasschieter, who works as an IT consultant at Capgemini.
He specializes in data and solution architectures for unstructured data. Reinoud has a deep-rooted experience in Enterprise Content Management. Overall, he is an expert in Artificial Intelligence with a specialised focus in Cognitive Computing.