Will AI help us or destroy us? So much drama! Generative AI will be the end of the world. AI is is a mass-delusion and making people feel “like they’re losing it.” As usual, the truth is somewhere in between. It will not save the world and probably will not destroy it either. Generative AI has benefits and risks. The best way to implement AI into your workplace is with eyes wide open and a thoughtful plan.
AI IN BRIEF: At the risk of oversimplifying, most of the new tools that are labeled as “generative AI” that people use are large language models (LLMs), which work by using statistical analysis to determine what is a probable response to a user’s query. ChatGPT, for example, is an LLM. LLMs need to be trained on huge amounts of data to be functional. The quality of the LLM’s output is based on the data used to train the LLM. (The phrase “Garbage in, garbage out “comes to mind.)
Worth noting: an LLM’s decision making is generally a black box – users see the input and the output, but they do not know why the LLM generated the answer that it did. Critically, LLMs do not reason and they do not conduct research. They generate text that looks plausible using data that has been fed into it.
THE PROBLEM: While AI can produce genuinely helpful and accurate information, it also makes AI prone to generating false information that is euphemistically called “hallucinations.”The most common examples of hallucinations that have hit the news are false citations to papers, books, articles and cases that do not exist or do not say what the LLM is quoting. Again, this is a product of the fact that the AI is not thinking, reasoning, or doing research – it is generating plausible looking text using statistical analysis.
THE ISSUES: Anyone in HR knows the pitfalls of using algorithms in their employment decisions. (The class action suit against Workday for age discrimination via AI use is winding its way through the courts.) But beyond HR functions, AI has enormous privacy, trade secret, copyright and security implications. Whatever is put out there loses copyright and other protections, including the private employee information. Generally, vendor contractors for AI absolve the platform from liability–it is on you.
THE LAW: Because the federal government has not regulated AI, the states are moving quickly to fill the gap. Colorado enacted a sweeping law that applies to both developers and users of AI, like the European Union does. California has several AI regulations. Illinois, Texas and Utah have amended or updated their AI laws. Massachusetts, Connecticut, Vermont, New York, Iowa, Minnesota, Nebraska, Oklahoma and Hawaii all have active bills that would impact AI. Without a comprehensive national approach, we will see a patchwork of AI obligations by state, making compliance trickier.
WHAT TO DO: Your workplace is probably using AI in some form or your employees could be– with or without–your knowledge. Businesses need to know the who, what, why and how for anyone using AI, To protect your organization, you should have an AI Acceptable Use Policy, an AI and Employment Decisions policy, and information for both management and the workforce to learn the benefits and risks of AI. The best policies are ones tailored to your workplace because they are easier to implement. The downside to model policies–I am looking at you ChatGPT– is they do not take into account the particular risks your workplace faces. Without a precise approach, the actual issues and types of uses can be missed altogether. (Did I mention “Garbage in, garbage out?”)
WE CAN HELP: Our firm has spent a substantial amount of time reading, researching and discussing AI for our clients (and ourselves). After a lot of thought, we have created a fixed fee service that will provide the policies, FAQs and employee rules your workplace needs to safely use AI. Our AI Consultation Service starts with an assessment of your workplace to identify potential opportunities and risks in using AI tools in your work and HR processes.
Contact us. We would love to alleviate any AI anxiety at your workplace.