The current boom in generative AI comes with a number of drawbacks and wider implications that should be carefully considered before engaging with genAI. 

These include:

• The exploitation of workers who are tasked with looking at traumatizing and illegal material in order to tag it for training data.
• The potential loss of jobs or underpayment of workers who have been told that AI can replace them
• Wasted energy and water
• Data centers, run in some cases by gas generators, that cause pollution in poor areas, worsening the health outcomes of those in poor areas
• Bias, reinforcing unjust systems in crucial aspects of life, including policing, sentencing, home loans, job applications, and more
• Increased surveillance and data brokering of personal information by the government and large corporations
• Government and corporate surveillance and privacy violations

There are a number of resources available to help users develop a deeper understanding of the broader issues and concerns that generative AI raises.

Big tech is betting big on generative AI as the future of business, a tool that will turbocharge efficiency and make many lower-level jobs, from gig artist to junior coder, redundant. However, generative AI also depends on an invisible underclass of low-paid laborers in the south to do the grunt work of building training data sets for LLMs. These workers are doing the hidden labor of collecting, viewing, annotating, and labeling data to ensure that LLMs don't, for instance, repeat the worst, most violently hateful things that appear in the raw data that comes from scraping the web.

The promised potential for generative AI to disrupt, devalue, and deskill human workers while also relying on an invisible underclass of human laborers makes generative AI a fraught field for fair labor practices.

While digital data has long been visualized as intangible and incorporeal, the truth is, data exists in servers that take up real land and run on real energy. Training LLMs is especially energy-intensive. As data centers proliferate in order to train ever-larger models, the environmental costs continue to grow with them. The demands of training and running large datasets increase energy consumption and worsen global warming.

Generative AI is only as good as its training data, and when its training data is the entire world wide web, it is prone to duplicating the same biases (in terms of race, gender, sexuality, class, and more) that are present on the web. This has serious real-world implications as governments and corporations outsource major decision-making to AI in areas of hiring, health insurance, policing, parole hearings, and more, with the likelihood that an AI trained on data from an inequitable society will make decisions that reinforce real world inequities.

The rise of AI has sweeping implications for privacy and surveillance, essentially creating a way for governments and corporations to sift through vast amounts of data they've collected. Even when AI cannot flawlessly navigate data, there's still the high likelihood of negative implications from governments and corporations acting on the belief that it is flawless.