#5 :: AI words + pictures

This week, the best chess player of all time, Magnus Carlsen, withdrew from a tournament for the first time in his career. This follows his July announcement that he will not defend his world title: “I simply feel that I don’t have a lot to gain, I don’t particularly like it”. His unease has been well documented. He’s become bored, jaded, by the precision and predictability of the game. Software has become so good that players can practise how the game will unfold beforehand, effectively playing based on memory.

We’re entering a new era. Computers and robots and AI are achieving better, more coherent results. Machines are using huge datasets, they are learning and identifying patterns, and deploy increasing computing power to generate their own. Driving cars and boats unassisted; diagnosing diseases; precision engineering; route planning; winning at chess… Computers are bringing us freedom: a new depth of human assistance. 

But what about in designing and creating, for imbuing deeper meaning in the world? Can machines replicate the unpredictability, serendipity, chaos of how humans generate ideas?

---

This issue is about tech and creativity and human-endeavour.

Collected this week:

1. Dall•E and AI generated art

2. Bowie’s Verbacizer

3. The painting robot

---

01: Dall•E and AI generated art 

My invite to join Open AI’s Dall•E mini beta arrived last week. Dall•E is a generative AI, capable of creating photorealisic images based on natural text input. What this means is that the user describes a scene in words - “a goldfish wearing a monocle, dining in the Gobe desert” - and the program creates a range of images to describe that vision. Here it is in action:

Above: Dall•E mini search and results

Dall•E, funded by Microsoft, is arriving at the same time as Google’s Imagen, and independent lab Midjourney. They are all generative AIs that produce images based on natural. They work by taking datapoints from billions of text–image pairings found online, and are suddenly getting very good. 

This is some of the most sophisticated AI around. What’s significant is the interplay of the human - the user - and the program. The computer doesn’t understand the nuance and meaning of the language. So, the input of the user is absolutely implicit — the complex semantics that people attach to images. 

Tech will continue to create new types of roles and skills for how humans interact with machines. Imagine how Dall•E might change the work of a graphic designer, a programmer, a copywriter, an artist… An artist won a contest this week based on a work first fashioned using Midjourney. And he’s keeping the text prompts he used schtum. 

This is interesting because:

1. Several platforms are emerging and becoming good at around the same time. The supportive tech has been in the making for some time but is now advancing at pace. So we’re suddenly arriving at an inflection point. Clusters and groups can then innovate and push the boundaries of the possible, and competitively push each other

2. Dall•E is designed for humans - to facilitate creative work. Computers are acquiring the traditional, ordinary skills of human labour, so humans have to now acquire next level, extraordinary skills…

3. As AI evolves, it may change jobs, making things easier, quicker, more productive, more efficient. But, as Magnus Carlsen asks, where is the fun of efficiency in creativity?

4. Dall•E can generate content for any purpose, and there are big ethical implications for deepfakes

---

02: Bowie’s Verbacizer

Of course, in the past, artists and writers and poets have always been developing and deploying their own low-fi AI techniques. Everything evolves from ideas that have come before. The Dadaists invented Cut Up technique in the 1920s, where written sentences were cut up and rearranged to make new sentences. These were then interpreted to listen for new significance and meaning. Later, William Burroughs relied on cut ups, believing that it tapped beyond the conscious, soothsaying things about the future, or insight about the past. A Western tarot.

David Bowie, always a pioneer of pushing tech into the creative process, digitised cut ups. In the 1990s, David co-developed an app called the Verbacizer, that ran on Mac. He would input sentences, and the software cut them up into new sentences. It gave him access to things he wasn’t thinking about, like a “technological dream”, that he used to generate lyrics and ideas.

Here he is in the 60s talking about cut ups …

And in 1997 about the creation of the Verbacizer …

This is interesting because:

1. The creative process is beautifully serendipitous. Can machines like the Verbacizer enable serendipity? Dall•E’s “Surprise me” button generates text prompts from which the user can springboard, like Google’s “I’m feeling lucky” randomised search button

2. AI is an augmentation, rather than a replacement, in the creative process. It should work in harmony with humans to innovate or generate ideas, like Bowie’s springboarded writing, or collaging Midjourney outputs together

3. We’re not going to have our jobs and creativity taken away by software in the future. We’re going to augment and facilitate and drive the creative process by ourselves, because we will be in control of computers. Humans will be curating the algorithms and datasets and outputs to make something new…

---

03: The painting robot

In Silicon Valley, the intersection of art and robotics is being explored by artist Agnieszka Pilat. She works with a giant, yellow robotic dog to swivel and stamp and swirl paint around her canvases. Spot the dog was originally created for the US Army, to detonate bombs and de-risk nuclear disasters. But now she has found a new profession as an abstract artist. Spot applies the paint, either by programming, or replicating Pilat’s movements in VR.

B70: Multiples, Agnieszka Pilat

Pilat has had residences at Google X and Boston Dynamics, worked with Airbnb and Tesla. She creates paintings of machines for the tech tycoons. Paintings made in collaboration with Spot have been sold by Sothebys, and hang in Virgin Galactic HQ. Her work suggests a post-contemporary, robotic era for fine art. Here is Spot in action:

Above: Courtesy of Agnieszka Pilat

This is interesting because:

1. In perfect harmony, Pilat creates the inputs and outputs of her system, and Spot executes the process. The artist and writer and musician and doctor and programmer isn’t replaced by AI. Pilat works in unison with the robot. Like with Dall•E and the Verbacizer, the human input is what brings sense and meaning to the creative process

1. But Pilat also talks about learning from the robot… it’s a mutual collaboration that creates something fresh

2. While AI is useful for speed and precision, Pilat’s methods can bring integrity and authenticity through the imperfection of the machine’s execution. And there’s a beauty in that…