AI Takes on Human Creativity

When it comes to Artificial Intelligence (AI), we immediately gravitate towards computation and efficiency. To date, AI has been limited to tasks which require brute force calculations, with the output being about accuracy beyond the capability of mere humans. AI, like machines of the industrial era, has been all about efficiency, using fewer resources and less time, to get the same thing done.

The narrative around this has been very human – it frees us up to get more creative. Creativity, art and nuance have, to this point, been where humans still shine. But that’s all about to change, as AI rapidly enters the realm of human creativity. And there will be significant investing implications to accompany this shift.

Art Imitates Life

When it comes to artwork and digital images, we have been using AI for as long as the graphical user interface has been around. But so far, the tools have only been able to copy, share, store and manipulate existing images. Art and new imagery have mostly been the job of humans. However, this year we’ve seen a quantum leap in what is possible when AI generated imagery, or dare I say “Art”.

Open AI projects have led to a number of art engines emerging, the most prominent being Dall-E-2, Midjourney and Stable Diffusion. The way they work is pretty much like a Google search does, except you’re instructing it to create something instead of instructing it to find something.

Just like Google, the AI engine trawls the web. But instead of serving up what it finds, it uses what it finds and then combines it with Deep Learning Algorithms to create something entirely new. These image-generating engines are able to combine concepts, attributes and styles.

The engines respond best with the use of natural language where the situations are clear and specific. To be sure, the results are weird, and often unlike what a human would create – probably a good thing.

To demonstrate, I used Dall-E-2 to generate some art pertaining to recent financial topics reported right here on the Eureka Report.

Prompt: “Cyberpunk Alan Kohler doing finance news report on a cryptocurrency winter”

Prompt: “$30 billion being burned to trying to convert Twitter into a super app”

Prompt: “The Cult of the Technology Founders in Silicon Valley”

You can see that the engine has some way to go when it comes to combining technology, news and finance. But when simpler prompts are put into the engine, from which it has more data to “learn from”, the images are not too far from what you’d expect:

Prompt: “Politicians drinking champagne during a climate disaster”

To me this is not that far off a cartoon you might see in the financial pages reporting on climate policy. And while a picture may say a thousand words, the word themselves, too, are now starting to be generated by deep learning algorithms.

Modern Scripture

AI generating text isn’t new. For years simple factual data has been used with pre-prepared scripts to structure informational reports. They’ve been used very successfully in the financial sector: For example:

“Today the All Ordinaries fell by 1.2% which was driven by lower commodities prices affecting the mining sector. While banking held firm as interest rate rises enhanced retail margins slightly on the big four”

Bloomberg has been using AI to create a constant stream of reporting across sectors for almost a decade. These simple AI’s have also been employed in sport reports where the key data is simple scores and statistics from a game.

But now the game has changed. AI can generate text with a far more complex syntax and descriptive text on pretty much any topic at all. It’s so good, it really is difficult to know it wasn’t written by a human. Again, this is driven by providing the AI with prompts.

Jasper AI is one such service. While it can’t quite write the Great American Novel (yet), it says it can create mind-blowing content ten times faster than any human in over 25 languages. For under $300 a year, Jasper claims the ability to write content for marketing, SEO, web, marketing copy, sales emails, video scripts and even love letters. Let’s find out! Below is an experiment to see how it may be able to help me write this article.

Prompts:

How will AI change the way people write Journalists want to know if this technology will replace them

Results:

“The advent of Artificial Intelligence (AI) has changed the way people write and consume content, particularly in the world of journalism. AI-enabled technologies such as automated text generators and natural language processing are now being used to create content for news websites and blogs, which is quickly changing the media industry…. As AI-generated content continues to improve, journalists must adapt and stay informed about advances in the technology….. In addition to using automation technology to write stories faster, AI can also help journalists understand their readers better. Natural language processing (NLP) algorithms enable them to analyze huge data sets of user behavior and determine what topics are popular at any given time or which types of phrases resonate with readers…. Overall, AI is making it easier than ever before for journalists to produce high-quality content quickly and efficiently while understanding their readers’ preferences better than ever before. As this technology continues to progress and become even more sophisticated over time, it will likely continue to revolutionize how we write…

Those two simple prompts above generated more than 700 words. I cherry-picked the bits which worked best – which is how the technology is supposed to be used. It took less than a second. Extraordinary.

Another thing the AI does is learn how you write, your turn of phrase, even humour, and like all AI it gets better the more you interact with it. In many ways this technology becomes personal – like a quasi-AI assistant.

There’s a further form of technology democratization which will emerge from this. The ability for every person to become a computer programmer. Within a few short years we’ll all be able to write our own software applications – literally write code – just by using natural human language. Microsoft’s GitHub CoPilot, which has been around a few years, offers programmers suggestions as they write code. GitHub claims that once coders enable CoPilot, they let it write 40 per cent of their code. It seems like autonomous code will take us a lot further than the promise of autonomous vehicles.

Next Era for AI

The first era of AI on the web was about organising and retrieving content, content which was made by humans, for humans. To this point all we’ve really built is digital repositories. The next era of AI is going to be about generating content for humans.

Search engines will be replaced by creation engines.

AI engines will create copy, visuals and eventually video on demand once the technologies cited above are combined. Potential Future Prompt: Create a 10-minute cyberpunk movie of what Sydney looks like in the year 2050.

The economic and investing impacts will be staggering. The cost of creative production will plummet and invent another era of disruption. What we need to look for as investors is which firms will become more profitable as they become more efficient and which firms will become redundant entirely.