Article: How You Can Break Into AI Without Computer Science

Photo Source: PxFuel

Artificial Intelligence (AI) and machine learning (ML) command attention today like never before, but the fallibility in these systems requires people to fact-check the answers. The market for AI-driven products continues to explode as large and small companies scramble to leverage the new capabilities of generative AI such as chatGPT. ChatGPT and other large language models, or LLMs, like Google’s Bard and Nvidia’s Megatron can generate document summaries, concoct recipes, churn out computer code, and write essays in seconds. These computer models use vast amounts of text and images from the World Wide Web as the basis of their systems, but these LLMs do not just look up answers to questions. Instead, these LLMs make guesses at the correct answers based on how people have written information on the web. It is a probability game, not a copy-and-paste program. Such systems have an uncanny ability to generate human-sounding writing, but the probability part of the model can also write convincing-sounding answers that are dead wrong. These types of answers are referred to as hallucinations. A spectacular example of a GPT hallucination occurred recently in Federal Court when a personal injury attorney named Steven A. Schwartz entered a court filing that cited six legal cases as precedent for his client against Avianca Airlines. However, the cited cases do not exist. Mr. Schwartz used GPT to write up his case, and the model wholly made up the cases. (forbes.com) Mr. Schwartz did not check the facts and now faces sanctions from the judge in the case. Because of the tendency of chatGPT to invent answers to questions, every company that wants to use this technology to power its chatbots or empower its employees to work faster must contend with the possibility of generating false information. For this reason, people with expertise and experience from all over the community can now play a role in fact-checking and testing the output of GPT and other models. People interested in AI/ML but who do not have a computer science degree can play a vital role in evaluating the integrity of LLM outputs. For example, large tech companies such as Apple advertise for positions in assessing the truthfulness of LLM outputs with the title “AIML-Annotation Analyst.” (jobs.apple.com) Such a role may require fluency in a language or skills in writing. The companies making new AI need real people to evaluate the output of their models, especially with the danger of hallucination-producing fake information. Companies such as Full Fact AI employ fact-checkers around the world to battle bad information about politicians and from news organizations and public institutions. (fullfact.org) A quick search finds other fact-checkers who have set up freelance businesses and offer their fact-checking skills through advertising online. Everyone with experience, from nurses to school teachers, police to auto mechanics, can make money from using their experience to identify AI hallucinations. The explosion of large language models such as OpenAI’s chatGPT has opened a vast array of opportunities for large and small companies to use this new technology. For all the promise of LLMs, this new technology has a dark side—hallucinations. The risk of false information from LLMs threatens businesses and has generated jobs for people who can evaluate these new tools for accuracy, truthfulness, bias, and security. If you are interested in working in the emergent world of generative AI, your experience and expertise can help protect us from harmful information and provide exciting work.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.