Background To AI
Introduction
Definition of AI is
"...machines acting to mimic human cognition to solve problems..."
Declan Foster, 2023
As humans we find it hard to visualise and analyse things beyond three-dimensional perspective (3-D); however, machines can visualise and analyse things way beyond 3-D, simultaneously.
However, AI is linear in thinking, ie straight line and logical, while humans are more lateral and unpredictable.
Some techniques used in AI:
- linear regression (is used to predict the value of a variable based on the value of another variable or variables, using a straight line; expresses the linear relationship between variables)

(source: Declan Foster, 2023)
- logical progression (next step(s) follows on rational)

(source: Declan Foster, 2023)
- clustering (groupings of similar characteristics or traits)

(source: Declan Foster, 2023)
- Deep learning (artificial neural networks - ANN)

(source: Declan Foster, 2023)
AI is the basis for ChatGPT; basic unit is a neuron or node which are connected and these analyse the data; can revisit the data for further analysis multiple times relatively very quickly, even hundreds of thousands of times.
ChatGPT and LLMs (large language models)
GPT stands for generative pre-trained transformer or generative artificial intelligence software. It is not an application; it is a platform.
ChatGPT started in late 2022 and had 1 million users in 5 days!
Microsoft is backing Open AI (developer and owner of ChatGPT) and is linking it with Bing; while Google has its own generative AI software called LaMDA. Currently, ChatGPT has the jump on Google because of using open-source software development.
"...to understand the disruptive powers ChatGPT, you have to compare what Google search delivers to customers with the results delivered by generating AI. Google search will deliver a list of links to websites and often the search results are driven by commercial advertising arrangements. When fully operational in Bing, someone searching on Bing with ChatGPT API installed will be able to see both links to web pages and generate AI results based on analysis of the pools of data that the AI machine has been trained on..."
Gilad Grinbaum as quoted by Tony Boyd, 2023
Some future uses of ChatGPT include
- challenging the delivery of news on smart phones
- replacing advertisers, telemarketers, analysts, software coders, authors, etc working in different formats
- customerising or personalising podcasts on almost any topic.
Potential uses of ChatGPT in change management include:
- as a memory jogger for tool and concepts like 8 steps in Kotter's change model
(for detail, see elsewhere in the Knowledge Base)
- providing questions for a role-play scenario, eg questions to ask a challenging stakeholder, resisting staff member, etc
- creating outlines via brainstorming and content for training courses
- summarising text, ie summarise a one-page project description into one paragraph
- answering questions on a specific change or project
- comparing project communication with organisation's corporate communication for style and tone.
NB To maximise effectiveness of ChatGPT, programming skills like coding are needed.
Need to be careful that do not give sensitive information to ChatGPT as it will become publicly available.
Some challenges that you need to be aware of when using degenerate if AI such as ChatGPT
- hallucinations (a predictor, ie it is just trying to predict the most likely next word in a sentence; it does this by searching through its available data; does not fact check itself)
- stochastic parrots (it is just repeating back words, phrases, etc, that it has in its storage; stochastic refers to deliberate randomness and it doesn't always produce the same answer to the same question)
- copyright (information, etc use without consent; as it doesn't disclose sources; the accuracy of the sources needs to be checked; can watermark AI-generated content, etc)
- bias (it was trained on larger data sets from the Internet, such as Wikipedia (as only 15% of Wikipedia users are female, it has a male gender bias built into it); very hard to remove bias owing to where it gets its information)
- ethics (need to reduce misinformation, hate speech, etc)
It is currently best used as a brainstorming tool to give you a greater range of possibilities, ideas, etc to select from.
Preparing your organisation of AI era of technological changes requires 6 elements to be considered.
Six elements (see below diagram)
i) AI organisational maturity model (like one from Microsoft, etc)
ii) helicopter view (engage with maximum number of stakeholders which are going to be impacted by the change; important to have a successful pilot project, etc)
iii) increased awareness and education (how to address the opportunity; use informal training sessions, etc; understanding the pros and cons of AI, etc)
iv) establish a governance framework (the shape and size depends upon what industry you are in and the size of your organisation; use of technology; ethics, etc)
v) appreciate your data (includes documenting lessons learnt, etc)
vi) operate as a learning organisation (understand double loop learning; be reflective; value feedback; encourage continual reviewing, monitoring and evaluating, and making the necessary modifications based on the findings, etc).

(source: Declan Foster, 2023)
New roles based in AI:
- trainers (teaching 'soft' (people) and 'hard' (technical) skills)
(for more detail, see elsewhere in the Knowledge Base
- explainers (main role is to bridge the gap between technologists and leaders)
- sustainers (ensure that AI systems are designed and operated as expected; handle unintended consequences; focus on ethics)
- prompt engineers (ensure that instructions are structured correctly; using appropriate prompts to get the right results).
Distinguishing between AI or human generated output
Already software has been produced, (eg AIcheatcheck) that
"...uses its own AI models to predict if the text has been written by a human or a machine based on clues such as word choice and sentence structure......readability..."
Tess Bennett, 2023