What Does It Mean to Use AI in Our Work?

‍

A recent article read: AI won an art contest, and artists are furious. The core argument is that robots are making art, and robots aren’t human, unlike the other artists. Is this a reasonable distinction? What does it mean for people and AI to co-exist and even compete at the same tasks, be it in contests, or for jobs?

In the legend of John Henry, the famed steel-driver competed against a steam-powered rock drilling machine. In the story, man beats machine, but his heart gives out from the effort. While the story itself is unverified, the tale stands tall on the edge of the industrial revolution as the last time man bested a machine.

No one would expect a person to outpace a modern jackhammer, which is why jackhammers are standard equipment in modern street crews. While lumberjacks may use chainsaws for their work, certain events at lumberjack contests are limited to axes and muscles, not power tools.

We use power tools, and even whole robots on assembly lines. Software engineers use tools to automatically generate rote code, instead of writing it by hand. AI tools are used to help marketers predict customer demand, doctors identify tumors, people compose emails (and articles–although this was written by hand), and help people with various work. Other than the occasional luddite, most people have embraced these tools because they removed the rote tasks.

What about art? Commercial visual media, from logos to movies, are now all made using software. Modern software can start to make recommendations. A simple version of this are the dynamic color generators which can recommend, using color theory, additional colors to add to an image giving the existing palette. PowerPoint has started to recommend different slide layouts. These will only get more sophisticated over time.

What about art for art’s sake? It’s hard to say where to draw the line. Consider, if there was a tool in which an artist would put her paintbrush, and it would decide if the brush needed to be cleaned and then automatically clean it, would we see the same level of outrage? I doubt it. Yet brush strokes are part of the art and art historians and students do look at the brush strokes of the masters. Software didn’t write this article, but it did spell check it and even gave me recommendations, such as replacing a phrase with a more concise one. Did AI help write this article?

You might say the tools just described are support, not the creative part of the process. It could be argued the software is just a modern paintbrush. But if the system makes any recommendation at all, then isn’t it a slippery slope?

And just who is making the recommendation? It’s the software, but the software was created by people, possibly using prior art to train the algorithm. How many artists, such as painters, sculptures, filmmakers, and writers, get input from others? How many get their inspiration from looking at prior great works? Software just automates this.

Jason M. Allen created the winning artwork "Théâtre D'opéra Spatial". The category he entered was for works that used "digital technology as part of the creative or presentation process." Those lumberjack contests that use axes also have categories that use chainsaws. In competitive chess, there are now human tournaments, software tournaments, and centaur tournaments. The later class was created by former world chess champion Garry Kasparov in which a human and computer together play a single side. The computer can out calculate a human in looking at possibilities, but the human head brings intuition.

Allen notes that it took him over eighty hours to create three pieces. He had to repeatedly tweak the input he gave to the software and created over 900 images during the process, and then used photoshop to further modify the images. That sounds like significant work. We can imagine, however, this workload will come down over time. Is there a minimum amount of time needed for it to be work?

There’s a bright line rule for the use of computers in some cases. In Las Vegas you cannot use an electronic computing device when gambling. My friends on the MIT Blackjack Team, however, didn’t need such a device, just a good memory and the ability to do quick calculations, which was not illegal. Some websites claim some casinos allow a calculator at the poker table. However, I’d imagine a fully on computer that reads cards played and dynamically calculates odds would be one step too far.

Even when computing devices are used, there’s a question of credit. Google translate got so good because it had access to lots of translations, as if it had been given millions of Rosetta Stones. Those Rosetta Stones had been created by tens of thousands of humans. Better jackhammers are created by engineers, who themselves do not make money jackhammering and their incomes aren’t harmed by the new tool. Translators make money translating. Their translations trained the AI that now replaces them, hurting future demand for their services.

This art was created because it had access to prior art. Some of that training art is now in the public domain. Some may not be. The law isn’t clear on the use of IP in training. What does it mean when people in an industry inadvertently train the software that replaces them?

On the other side of it, the artwork itself is public domain since it was created by an algorithm. If I use the blur function on a drawing tool, that too, is the application of an algorithm. Where is the line where we cross from algorithm assisted to algorithm generated?

More generally, what obligations, legal and ethical, do we, as a society, employer, or engineer, have to those who sow the seeds of their replacement. These are the questions we need to answer. Not having answers doesn’t make them go away, it just makes the future more volatile and unpredictable.

Before you dismiss this as an edge case, and unlikely to impact you, know that the future is closer than you think: A Chinese game company has appointed the world’s first humanoid robot as its CEO.

‍