Advanced AI Development Should be Paused
Halting development of advanced AI
In January I wrote about the potential danger of advanced AI with the introduction of ChatGPT. Now, several luminaries of big tech, and social and philosophical disciplines are requesting a temporary halt of six months to the further development of advanced AI systems more powerful than ChatGPT-4. There is an open letter crafted by the Future of Life Institute that has been signed by the likes of Steve Wozniak, one of the original founders of Apple, Yuval Noah Harari, noted historian and philosopher, and Emad Mostaque, CEO of Stability AI, the founder of the AI image generator, Stable Diffusion. The signers of the open letter believe, and I agree, that we must better understand the impact of this new technology on humanity before continuing development. In the letter several requests are being made of the advanced AI development community. Among them is that there be shared safety protocols developed for any future designs and that advanced AI developers work with policymakers to create AI governance systems and establish strong regulatory bodies. Emad Mostaque, one of the signers of the open letter, is currently in court because his AI product, Stable Diffusion, was trained on the work of artists who had not given the company permission to use their work in that way and were not compensated.
The open letter sounds like a step in the right direction. Social media was allowed to develop with very little oversight. It was making money for the platform developers and company shareholders through advertising and very little attention was given to issues of data collection, targeting, privacy, and age appropriateness. It is good that we are attempting to address issues of potential big tech disruptions with advanced AI as it is being developed.
Filing suit against AI image generators
AI image generators are being challenged in court by artists whose work the intelligence was trained on. The artists did not give permission for their images to be included in the training data and the artwork is not in the public domain. They were never asked to participate. The courts will be weighing in on the issue of copyright infringement and if the images that the AI comes up with are derivative works. Who owns the copyright of the AI generated image? Is it the person that created the prompts given to the AI to compile the image or is it the AI company or is it the person whose artwork the AI references to paste together a requested composite image? Is AI creating anything original that can be copyrighted? How will or should artists be compensated? The compensation of artists is key. I find it hard to believe that Stability did not try to negotiate licensing agreements with any of the creators whose artwork they trained their AI tool on.
Adequate compensation is already a problem in the visual arts industry. So much content is available online digitally and is literally there for the taking. Respect for the rights of the creator diminishes with each copy and paste command. The visual arts have already suffered a major disruption by the introduction of digital tools to create content and that has both lowered the point of entry into the field and reduced fees. I think that many of us recognize that the genie is already out of the bottle, since the genie depends on the content of other artists to function, then the artists should be adequately compensated, and their contribution noted. AI based image generation will further narrow job opportunities that are already shrinking. Karla Ortiz, a concept artist, who is one of three artists suing Stability AI asked, "Why would someone hire someone when they can just get something that's 'good enough'?" when interviewed on CBS Sunday Morning. Not only do you get something that is good enough, but you also don’t have to pay for the work or pay as much. The music industry is not immune to this problem either.
We tend to conflate speed and efficiency with productivity. The more tasks you “accomplish,” the more productive you are. You are seen as valuable in the eyes of society and your workplace. There are problems with this view point. The productivity curve for the American worker has climbed substantially, but compensation for these gains in terms of wages or leisure time are stagnant. The productivity per American worker has increased 434% since 1950. One way to look at that is that it should take less than one-quarter the work hours, or less than 10 hours per week, to afford the same standard of living as a worker in 1950 (in other words: our standard of living should be over 4 times higher than it is), but it is not. People are working longer hours with little or no shut off switch because of mobile devices. You are not compensated for the extra hours you put in. Wages are stagnant. According to the Economic Policy Institute, real median hourly wages for the U.S. worker has essentially been unchanged from 1979 to 2019. The introduction of computers in the workplace has made workers more “productive,” but the cost is high. Technology has sped up the workplace environment to the point where it outpaces the capabilities of human beings.