Go back ten years and you’ll find hours of video, countless articles, and hundreds of podcasts with AI experts explaining why creative jobs will be the last to be taken by artificial intelligence. It now seems that AI can be just as, if not more creative than most humans. Why were the experts so wrong?
AI has made huge strides in creative fields in recent years, surpassing expectations made just a few years ago. While it’s long been used for tasks requiring high precision and accuracy, such as data analysis and image recognition, it’s now also being used to generate creative content in other fields.
With the advent of AI tools like ChatGPT and DALL-E 2 from OpenAI, and other image generation tools like Stable Diffusion or MidJourney, it’s now becoming clear that creativity (at least how many would define it) is not something that’s solely unique to humans and animals.
What is Creativity?
The Oxford English Dictionary defines creativity as:
“The use of imagination or original ideas to create something”
Clearly (at least I hope), AI doesn’t have an imagination, nor does it have “ideas”, at least in the sense we usually understand them. So it’s not creative, right?
The problem seems to be that most definitions mention “thoughts” or “ideas” as part of creativity, and we like to think that AI doesn’t have them; both require a mind.
When I asked ChatGPT to define creativity, it said:
“Creativity is the ability to generate new and original ideas or to find new and innovative ways of solving problems. Creativity involves both divergent processing, which involves generating many different possible solutions, and convergent processing, which involves evaluating and selecting the best solution.”
As you can see, one of its major drawbacks at the moment is that it tends to use 50 words when ten will do. But despite this, the definition above sounds as good a definition to me as any that might be given by a human and doesn’t involve thoughts or ideas.
Whether AI is really creative is largely dependent on who you ask. If who you ask takes creativity to involve thoughts or ideas, then it doesn’t. If they believe that creativity is more based on the end product or solution and doesn’t require a biological brain, then they likely believe that AI can indeed be creative.
Why is AI so “Creative”?
Perhaps the main reason AI is so good at appearing creative is its access to vast amounts of data combined with its ability to process information and identify patterns within it much faster than any human.
An AI system can analyze thousands of paintings by a particular artist and identify common themes or techniques that the artist used. Not so much in art, but in vast data sets, these themes might not even be apparent to humans. It can then recreate them in this style based on user input.
AI can appear creative as it can be programmed to follow specific rules or guidelines. An algorithm could be trained to generate music that follows certain melodies or chord progressions. It could then create a new composition that adheres to certain conventions but still has a unique sound.
The entire basis of reinforcement learning, a key area of AI, is that it can learn from its mistakes and improve over time. If an AI system is used to generate poetry and the resulting poems are not well received, the system can learn from this feedback and adjust the output accordingly.
This ability to learn and adapt makes AI particularly well-suited for tasks that require a high level of creativity.
Examples of “Creative” AI
There are countless examples of AI showing elements of creativity. Here we’ll go into some of the most exciting and impactful cases.
Art
In 2022 controversy erupted in the art world when AI-generated art began winning competitions. Many believe that AI art should have a separate competition and shouldn’t be allowed to supersede the achievements of human artists, despite its quality.
In the example above, you can see that the Colorado State Fair (won by an AI artwork in 2022) believes the evolution of AI art generators helps the fair evolve and embrace important conversations around innovation. If the AI artwork is better than a human artist, should the AI be banned, or should human artists be responsible for finding ways to improve their own work?
Go
The brilliant AlphaGo documentary follows the development of DeepMind and AlphaGo in their quest to beat the world's best Go player. The most famous move in the final games is known simply as “move 37”.
People with a much greater understanding of Go than me say the move clearly exhibited “original thinking” and was a move a human would never think to make. Lee Sodol, the AI’s opposition and widely regarded as the best Go player ever, repeatedly described the move as “so so beautiful.” AlphaGo went on to win the five game series 4-1.
It’s incredible that an AI can impress people at the cutting edge of their field this much. It seems at least close to something like creativity to pull off a move so unexpected and yet so compelling and original.
Literature
ChatGPT seems very creative and can write almost anything you can think of. Unfortunately (or perhaps fortunately, depending on your perspective), in most cases, it can’t do it as well as we can. It’s particularly bad at making text feel human, often repeating the same words and phrases, and the text it generates is very dry. It also has a tendency to state something confidently but be totally incorrect and never cites its sources (unless you use this Chrome extension).
These examples clearly show something resembling creativity; whether they really show it or not is debatable. But why did so many people say that AI could never do anything even close to creativity?
Why were the experts so wrong?
In just 2019, John Abel, vice president of cloud and innovation at Oracle, said:
“As we know with the modernization of IT, and specifically with AI, machine learning, anything that’s a logical processing job will be at some point replaced, so what we’re asking our staff to do is use their creative skills across all age groups (in) the workplace because that’s the unique advantage.”
In the same year, John Fallon, CEO of Pearson — the world’s largest educational publisher — said the capabilities most insulated from automation were: “uniquely human skills like creativity, persuasion, and empathy.” As we’ve seen, AI is already incredibly creative and can be persuasive as well, at least in text format. Maybe one day, it’ll be able to show real signs of empathy, much like the AI in Ex-Machina (well worth watching if you haven’t seen it).
Martin Ford, a futurist and author of Rise of the Robots: Technology and the Threat of a Jobless Future, said in 2016 jobs that require “genuine creativity, such as being an artist, being a scientist, developing a new business strategy” are the most resilient to automation. Yet ChatGPT can write a business strategy, many AI programs can create art, and DeepMind solved one of biology's grand challenges - protein folding, using AI.
These people were clearly at the cutting edge of AI development, so how did they get it wrong?
I think the main reason is that they believed creativity was something only biological life could exhibit. Creativity is often viewed as something that makes us human. It feels unnerving somehow to think that a machine can be just as, if not more, creative than us.
It could also have been that they found it hard to imagine the growth that took place in the AI space and the effect it would have on the creative aspect of AI. Exponential growth is notoriously difficult to predict, and thanks to things like Moore’s Law and the availability of data, AI has been able to take full advantage of it to improve as fast as it has done.
Let me know in the comments below if you have different ideas. Would love to hear them!
What does it mean for the future?
Perhaps AI will be able to do everything just as well, if not better than humans one day, including creative tasks. I think this is likely on a long enough time horizon, but even if it can do it “better,” people still enjoy seeing artwork made by a human and thinking about the effort that went into it.
There’s AI today that can be any human at chess, but we don’t stop playing it. The best chess players these days are a combination of human + AI, and this could be the future of creative tasks. A human might create the art, and AI might be used to finetune it, or an AI could provide a brilliant outline of an article for a human to write in a more empathetic manner.
The protein folding example above shows how creative AI may benefit humanity and how it can process information differently to humans. This is almost certainly just the first of many advances that will be made in scientific fields in the next decade - I’d bet that from now on, nearly all will be made with major help from AI.
With advances in virtual reality, perhaps the future will be one lived partly in a world created by AI and presented to us using technologies like AR/VR to create a utopian (or dystopian depending on your view of AI) form of existence.
Maybe AI can never be truly creative because it can never produce something totally novel with no real-world data to be trained on.
But does it matter? If it appears just as creative as us, will anyone care that it’s not exhibiting creativity as we normally think of it in the form of “thoughts and ideas”?
I’ll end with a quote from 1964 by perhaps the most famous science fiction author ever - Isaac Asimov, to show just how difficult predicting the future is. In one sense he was right; psychiatry is, arguably, the most important medical field today, but not for the reason he thought. In fact, it’s almost the exact opposite - we have too much to do, but will he be right given another 10 years?
“Even so, mankind will suffer badly from the disease of boredom, a disease spreading more widely each year and growing in intensity. This will have serious mental, emotional and sociological consequences, and I dare say that psychiatry will be far and away the most important medical specialty in 2014.
The lucky few who can be involved in creative work of any sort will be the true elite of mankind, for they alone will do more than serve a machine.”
Great article, interesting stuff!