An Artificial Intelligence Chatbot Primer
AI Terms and Definitions and Chatbot app terminology.
The ‘Artificial Intelligence (AI) chatbot writing revolution’ has infiltrated our lives. By now, you have no doubt heard of apps such as ChatGPT or Bard. It’s also likely you’ve experimented with AI-generated query response apps to see what they can accomplish.*
You might even be using them on a daily basis, depending on your job role.
*Common uses for AI-generating apps like ChatGPT or Bard include content creation or curation, policy drafting, email drafts, topic summaries, and other uses.
History of AI chatbots (AI-generating technologies)
AI chatbot apps have been around for a long time. The difference is that AI-response-generating technologies have improved significantly in recent months. Newer chatbot apps like ChatGPT or Bard use larger datasets (large language models/LLMs), faster processing speeds, and more sophisticated computer programming sciences such as algorithms.
Compared to modern apps, earlier chatbots had the following weaknesses:
- They ran on much smaller datasets
- They used rudimentary “if/then” programming and less sophisticated algorithms
- They offered limited responses to keywords/key phrases
- Nor were they adept at ‘self-improvement’ when they responded erroneously to an input or request
- For example, when someone typed the words “‘get me a real human” in ALL CAPS, or
- Screamed out ‘Get me a person, @#S&%#, not a bot!’ in an amplified voice
Today’s AI-driven chatbot apps, e.g. ChatGPT and Bard, are far more capable of responding to questions in nuanced, human-like ways. They offer powerful, time-saving solutions for a variety of industries and job roles. But they remain in an ‘experimental’ stage of development and use.
Source: OpenAI ChatGPT, Reuters, et al.
AI-Generating Apps: Use Trends & Adoption Curves
- The fact that you and millions of other computer users are likely already using ChatGPT or Bard on a near-daily basis is impressive.
- That’s because these apps were only recently released to the computer-using public.
We are currently only six (6) months into the release of ChatGPT and Bard to the broader public. The ‘early adoption’ tech curve is on the steepest incline ever witnessed.
GMP Industry Warning on AI Software Uses
Reminder: If you work in a regulated industry (such as pharmaceutical manufacturing, medical devices or medical software production, therapeutic goods warehousing and distribution, medicine labelling and packaging, etc.):
- You will need to conduct and document a risk analysis for the AI-driven software
- You will need to implement data integrity assurances relevant to the intended use of the software or its outputs (AI-generated query responses)
- You’ll need to ensure your employee’s use of AI-driven software meets regulatory requirements for computerised systems used to perform or enhance GMP activities
Current Regulatory compliance requirements for GxP/GMP computerised systems (including software components/algorithms) include:
- Data integrity requirements (GMP guidance for data governance)
- The FDA’s Computer Software Assurance (CSA) guidelines
- GxP Computerised Systems guidance as detailed in ISPE’s GAMP5 (2nd edition)
- PIC/S Annex 11
- Good Recordkeeping Practice (GRK) and Good Distribution Practice (GDP) regulations
- Software as a Medical Device (SaMD) regulations
- FDA requirements for Clinical Trial computer systems/data storage
- Other relevant guidance for batch records, computer systems, and cybersecurity measures
You’ll want to keep an eye on the ‘AI software regulations’ space, of course. Particularly as AI-driven apps are increasingly being used in the manufacture of therapeutic goods, as well as in healthcare settings, hospitals, and radiology departments.
Who uses AI-generating apps like ChatGPT and Bard?
Computer users of all skill levels are adopting chatbot apps, faster than any other previously launched application or tech device in computer-science history. The capacity of AI-generating apps to improve productivity is incredible, exciting, and frightening all at once.
So if you want to learn more about the landscape of AI-generating ‘chatbot’ apps, and learn some basic AI terms and definitions, this article series will get you up to speed.
Trends Analysis for AI-driven Chatbot Apps
Rapid, widespread adoption is something most software developers dream of achieving. In the case of AI-driven chatbots, however, it has left quite a few developers with AI-driven nightmares.
The tech-adoption curves for new apps like ChatGPT and Bard were unprecedented. These apps were only officially launched earlier this year (2023); although ChatGPT was ‘soft-launched’ late in 2022. Consumer adoption of these powerful AI-driven apps was so rapid and widespread it left consumers in awe and businesses, educators and governments in fear.
Expectations for AI-use Regulations and Company Policies
- Technological advances nearly always precede risk management controls such as regulatory guidance and organisational policies.
- Such is the case for regulations and policies regarding AI-generating query response applications such as ChatGPT or Bard; regulations and policies for AI chatbot use either don’t exist or are currently being drafted.
Consumers are raving about what AI-response-generating apps can accomplish. Meanwhile, software developers, thought-leaders, businesses, governments, and education providers are wondering how to manage its risks. A bit late, perhaps. The horse has bolted.
Even so, new regulations covering the legal and ethical uses of AI software technologies (including ChatGPT and Bard) are in the works. Expect to see them soon (ish).
What are AI-generating/chatbot apps?
- These apps were designed with sophisticated algorithms, massive datasets, and self-improving AI technologies.
- AI-driven technologies like ChatGPT and Bard can generate responses to queries in a matter of milliseconds (perhaps microseconds).
Many of these AI-generating query response apps will ‘write for you’ when asked properly.
- By properly, I mean phrasing your request in a way these supersmart software programs will understand.
- And by ‘writing for you,’ these apps can summarise just about anything, providing you with detailed, well-structured paragraphs likely to be grammatically correct.
Questions are usually typed into a dialogue box via a cloud-based app (internet access is required). Some query response apps also respond to voice commands, image uploads, or other ‘query’ prompts.
For example, you can ask ChatGPT or Bard to generate (write) a research summary, draft an outline for your University paper or write the entire paper if permitted by the University’s policy on AI chatbot use…although this is highly unlikely!
It can even generate themed party invitation ideas, write a cover letter or rejection letter…just about anything you ask it to do.
Example of ChatGPT queries (chatbot questions)
How to word questions for AI-generated responses
Here are some example questions you can ask OpenAI’s ChatGPT. Note: the free version of ChatGPT relies on a dataset that only includes data up to 2021. We’ll discuss such limitations further in Article 2 of this series.
Why are AI-generating/chatbot apps so popular?
Chatbot apps like ChatGPT and Bard are popular because business professionals, university students, and other consumers are bedazzled by what these apps can accomplish – and how quickly. They can generate well-formed paragraphs or other content in response to written queries or other inputs – in a matter of milliseconds. Let’s make that microseconds.
The outputs [of AI-driven query-response apps] indicate relatively few, if any, limits. From generating new songs and artworks to drafting your thesis, to AI chatbots passing a radiology medical exam, AI-driven apps are both powerful and diverse.
And that’s what makes it scary.
How are AI chatbot apps different to search engines like Google or Bing?
AI-driven query response apps synthesize information for you, in well-formed paragraphs or image selections, in response to your query inputs. Search engines, on the other hand, generate a bunch of links, sponsored content (advertisements), and snippets, which may or may not be useful to you.
So instead of sponsored content, snippets or web page links, query response apps can can generate detailed paragraphs and topic summaries specific to your questions or commands.
- There are also pros and cons to ChatGPT, Bard, and search engines including Google.
- This is true for all computer science technologies and new developments.
- We’ll discuss the pros and cons of AI chatbot apps, and how they are similar or different to established search engines, later in this article series.
Definition of a ‘chatbot’
Artificial Intelligence (AI) chatbot technology = AI-generated query response software/apps
A ‘chatbot’ is a computerised program (software or app) designed to simulate conversation with human users, over the internet. Early chatbot technologies were generally used on company and government websites. They were programmed to automatically “answer” commonly posed questions from an organisation’s client base or prospective customers.
- Chatbot technology has mostly been used to answer questions posed by website visitors, on a 24/7 basis (without the need for additional personnel or any personnel at all).
- Chatbots are often routinely deployed to reduce labour costs and/or improve efficiency.
- They can be useful (or equally useless) on shopping websites or during service outages, supply chain issue updates, widespread billing errors, etc..
They are not necessarily useful for good customer service, however. These chatbots generally failed to recognise the crux of a problem and/or failed to provide the appropriate client-desired solution.
Most consumers found early-version website chatbots to be frustrating, restrictive, and inept.
- But new chatbox AI technology is likely to generate warmer, quasi-relevant replies that sound like humans.
- These AI-generated responses may include problem recognition and empathic language about that problem.
Whether new AI-generating response apps will improve customer service ratings, or be able to offer client-desired solutions, is yet to be seen.
The real benefit of using newer AI chatbot technologies is in the context of having a ‘data savvy digital assistant’. An assistant you don’t have to provide coffee for. An assistant who doesn’t get cranky when you ask it for a task at ten minutes before 5 pm on a Friday. One on call for you, day or night, to help you accomplish a task or writing assignment on a very tight deadline.
AI-chatbot industry ‘hype’
Despite concerns about the safety and accuracy of AI-driven apps/chatbots and their impact on employment prospects & job skills, these apps are more ‘evolutionary’ than ‘revolutionary’.
Word-of-mouth sharing about how these apps ‘work’ was enthusiastic and swift.
- They are transforming our lives, primarily, in positive ways (although not always).
- Yet no technology (AI-driven or otherwise) comes without risks.
We’ll discuss the impact of these innovations on career skills, and other AI app safety risks, in Article 2 of this series. First, let’s explore AI terms and definitions relating to query-response-generating apps, including ChatGPT and Bard.
AI Terms and Definitions (Chatbot Terms)
1. What is Artificial Intelligence (AI) in the computer sciences?
AI stands for artificial intelligence (AI), a computer science field. Definitions of Artificial Intelligence (AI) technologies include concepts such as ‘machine learning’ or ‘deep learning’, datasets, and algorithms.
Advanced algorithms can generally generate responses to questions (queries) a lot faster than humans can, and these responses are often more accurate or use better grammar – but not always).
Algorithm (AI definition): The term algorithm represents a sophisticated mathematical equation (a coded/programmed component of a computer program) that helps answer a question and/or solve a problem.
- AI-driven technologies can be thought of as computer software programs/apps that use “mathematical models of data” to help the program or apps “learn, without direct instruction from a human” (Source: Cambridge)
- In simpler terms, these computer programs (apps and devices) are ‘smart’ technologies; and they are programmed to enhance their ‘smartness’ over time.
2. What does the term ‘AI-generating’ chatbot apps mean?
AI-driven chatbots like ChatGPT and Bard describe software/applications (apps) intended to produce relevant outputs by combining:
- Sophisticated programming that can mimic human language responses to questions (queries)
- Analytical capabilities/algorithms and parameters
- Extensive datasets (very large databases)
They can enhance human performance with minimal oversight – a Supervisor’s dream assistant. But AI chatbots can also get things ‘wrong’. Very wrong*. And they are not always as ‘self-reflective (or self-improving) as a person likes.
*Even dead wrong, as detailed in their disclaimers and ‘use terms & conditions consent waivers’.
- Checking AI-generated content before use remains critical to managing AI app risks.
- So oversight (checking, editing, and tweaking) of AI-generated output is not necessarily going to be minimal.
AI terms for content-generating apps, such as ChatGPT and Bard, include ‘query response generators’ or ‘language generators’, ‘AI chatbots’ and ‘AI-generating applications’.
3. What is the legal definition of AI?
The Legal definition of AI (artificial intelligence), as published by Cornell University with data sourced from US Code § 2358 notes, is as follows:
AI is “Any artificial system that performs tasks under varying and unpredictable circumstances without significant human oversight, or that can learn from experience and improve performance when exposed to data sets.”
4. Which job roles & industries will benefit most from AI-response-generating apps?
The time-saving, productivity-enhancing aspects of “query response generating” apps like ChatGPT and Bard will depend on numerous factors. Productivity influences for AI-generating response apps/chatbots include:
- How adept the app is at generating relevant and correct responses to queries
- Algorithms and datasets
- Weak points
- How adept the user (querist) is at query-writing or parameter-setting
- Computer skills and language skills (‘input’ skill sets)
- Ability to phrase questions that generate relevant responses (outputs)
- Experience with AI technologies including search engines
- The industry and intended uses for AI-generated information
- AI-tech user’s knowledge directly relevant to the topic being queried and the intended use of AI-generated information
5. How much time can you save by using ChatGPT or Bard?
Examples from respected trend analysts, such as Statista and Hubspot, along with my personal experiences, suggest individuals with marketing, sales and writing roles may be able to save 2 to 3 hours each day by deploying AI chatbot technologies.
That is, so long as they deploy these technologies in an effective manner.
When a professional and/or sales rep can gain (save) an extra 10 hours of time each week, every week of the year* by using apps like ChatGPT, imagine the potential productivity gains. Imagine the potential sales income increases!
*So long as the internet is working.
(And imagine...what could possibly go wrong? Which we’ll cover in Article 2).
6. Differences & Similarities between Bard and ChatGPT versus Search Engines
- AI-generating apps like ChatGPT and Bard have some similarities to browser searches.
- However, they are notably different to searching for information using a browser app/search engine such as Google.
- The answer to “Which is better, ChatGPT or Bard, or a Search Engine like Google or Bing?” depends on:
- Your intention for AI-generated information
- Your usage plans (what you plan to accomplish with AI-response queries)
- What you already know about the query topic
- How skilled you are at optimising certain types of algorithms
ChatGPT, Bard and similar Ai-generating apps are generally more efficient at answering questions and generating information than simply searching in a web browser. They are not necessarily accurate (more on this later). But they do offer efficiency gains.
- When using web browsers/search engines, you generally have to:
- Search for (sift through, and visit) various sources of information (web pages) individually
- Choose between (or ‘weed out’) sponsored answers to search terms (advertisements)
- Read/absorb long texts or web page links from websites with unknown reliability
- Use your brain and writing skills to paraphrase, summarise or tweak your discoveries
- When using AI-generating ‘query response apps:
- The software can gather, synthesise, and rephrase information from a very wide variety of sources, somewhat instantaneously
- It may sometimes arrive in a ‘nearly ready to use for content’ state (emphasis on ‘sometimes’)
- You will still need to use your brain, your subject knowledge, and your writing or editing skills to tweak the content to suit your intended purpose
- You may also need to dig further (and/or use a search engine like Google or Bing) to verify the information source and information reliability
Summary of Article 1 of 2 Articles in the blog series covering AI-Terms & Definitions and Adoption Trends of Chatbot Apps.
In summary, you may be able to use AI-generated content without spending a lot of time paraphrasing the query outputs or conducting further research. These apps often do a fairly good job summarising information in response to well-structured questions and other inputs.
But, as you’ll soon discover, the output likely still “needs work.
Sometimes, it will need a lot of work.
And sometimes, AI-generated output will be so erroneous it will need to be thrown in the garbage bin (i.e. deleted); or the AI app asked to ‘start over’ and generate a new response. Which it can, and will do, quickly, without complaining as a human might, when asked for a ‘do overs’).
It is true that AI-generated content might be entirely useless at times.
You will always need to verify the accuracy and appropriateness of AI-generated content in relation to your intended purposes.
Powerful but not perfect.
Finally, if you are already using these apps, you’ll recognise AI-driven chatbot apps are powerful, but not perfect. And if you haven’t yet explored what AI-generated response apps can accomplish? It is highly likely that you’ll do so in the future.
We’ll discuss the risks of AI-generating apps/AI-generated response technologies in Article 2, Risks of AI Apps including ChatGPT and Bard, published in June 2023.
Disclaimer: This article series was written by a human.
Further reading about regulations for AI chatbot apps like ChatGPT and Bard – Scientific American journal article.
Last updated on July 4th, 2023 at 11:17 am