Generative Artificial Intelligence and Chatbots (Some Insight, and  Two “Articles” That Don’t Exist)

Generative Artificial Intelligence (chatbots) is booming. There are opportunities and pitfalls for food fraud prevention. For example, the predictions are based on currently available information. Also, sometimes, the conclusions or results are incorrect, such as a chatbot providing me with citations for two articles that did not exist.

[Note: For convenience, I will refer to all generative artificial intelligence systems as ‘chatbots’ even though there is a difference between generative AI and the application that is conversational.]

This blog post idea started when I used AI in a chatbot as a background research tool to see if I missed any key articles or concepts. I found two amazing new articles – GREAT!!! These two articles had amazing titles that exactly met my research needs. But as I dug deeper, I found that the articles did not exist.

While doing this research about a year ago, I searched for the titles of the (non-existent) articles – no luck. I searched by the author – no luck. I then searched in the cited journals – no luck. I had reviewer access to one of the journals, so I could even search for articles that were under review and not public yet – no luck. Hmm… the article did not exist.

And then, last week, it happened again. A chatbot generative AI system gave me an article that did not exist.

This has such great promise, including for food fraud prevention, but caution is needed.  So, let’s review some details.

Artificial Intelligence vs. Machine Learning

Artificial intelligence (IA) is all the rage right now. It is so hot that it seems some of the industry consultants changed their job titles to Artificial Intelligence Manager from Blockchain Manager. Before that, they changed from big data to RFID or others. But what actually is AI, and how can it help in Food Fraud Prevention?

To be clear, AI will definitely be able to help in food fraud prevention, but the question is how and how much.

• Artificial Intelligence (AI) (from NIST): “A branch of computer science devoted to developing data processing systems that performs functions normally associated with human intelligence, such as reasoning, learning, and self-improvement.”
• Machine Learning (ML) (from NIST): “a field within artificial intelligence, focuses on the ability of computers to learn from provided data without being explicitly programmed for a particular task.”

Then the applications are:

• Adversarial Machine Learning (AML) (from NIST): “is the process of extracting information about the behavior and characteristics of an ML system and/or learning how to manipulate the inputs into an ML system in order to obtain a preferred outcome.”
• Generative Artificial Intelligence (from US GAO): “is a technology that can create content, including text, images, audio, or video, when prompted by a user. Generative AI systems create responses using algorithms that are trained often on open-source information, such as text and images from the internet. However, generative AI systems are not cognitive and lack human judgment.”
• Chatbot (from NIST) [Note: not to be confused with ChatGPT which is a brand name of a generative AI product] “AI-enabled chatbots use natural language processing models (NLP) to process and respond to human input, but these chatbots have more complicated architectures than just a language model (participate in a conversation).”
• Bot (from Webster’s Dictionary): “short for computer robot, a computer program that performs automatic repetitive tasks.”

Also, related terms are:

• Blockchains (from NIST): “are distributed digital ledgers of cryptographically signed transactions that are grouped into blocks. Each block is cryptographically linked to the previous one (making it tamper evident) after validation and undergoing a consensus decision.” [Blockchain network: “the network in which a blockchain is being used.”]
• Big Data (from NIST): “is a term used to describe the large amount of data in the networked, digitized, sensor-laden, information-driven world. The growth of data is outpacing scientific and technological advances in data analytics.” And “consists of extensive datasets primarily in the characteristics of volume, variety, velocity, and/or variability that require a scalable architecture for efficient storage, manipulation, and analysis.”
• Data science (from NIST): “is the methodology for the synthesis of useful knowledge directly from data through a process of discovery or of hypothesis formulation and hypothesis testing.”

So, how can chatbots contribute to food fraud prevention or food safety? It depends on the problem you are trying to solve and the data that you have available. I think the biggest key point in the definitions is “useful knowledge directly from the data.”

Chatbots to Help Food Fraud Prevention Research

A key is to define what AI/ML can provide and what task you are conducting. AI has been an excellent tool for me to support research, especially in the information-gathering stage. For example, when researching a term’s definition, I will often ask the same question to the bot. It is interesting to see what the bot finds and their conclusions. I find that if I’ve conducted comprehensive research then the bot doesn’t provide any new definitions and citations.

As a starting point, a chatbot is a great way to start researching a topic before going through more direct internet keyword searches. Over time, you will get more efficient at creating the prompts to ask the chatbot (Sidenote: ‘Prompt Engineer’ is a new job title that works for both the users and the providers.)

To review the accuracy and application of the searches, I asked a chatbot questions about my research area and about my articles. As would be expected, as my questions (prompts) got more detailed, the chatbot started to miss more details.

While chatbot-type systems take internet searches to the next level of efficiency, they still need to be reviewed. The conclusions are based on a wide range of sources that may be outdated, applied incorrectly to your question, or just be wrong.