I miss writing and lately Artificial Intelligence (AI) is an unavoidable subject that is triggering my annoyance at people. And it triggers it in the same way that anything that is new and exciting in technology and communication has triggered me since school.
AI is an amazing tool and it does open a world of possibilities for how we build things, how we communicate, how we speed up work. It’s not magical, and it is fertile ground for swindlers and peackocks who boast about their extensive knowledge of this bleeding edge field.
AI It isn’t really bleeding edge, it’s been a part of academia for years. What’s new is that the barrier to access AI tools came down. The same way that blogs brought down the barrier to publish content online.
My rant stops here, what comes next is my view on how and why we should look at AI as a tool and not a solution. Then we can imagine some scenarios for the way this new tool will change the way we communicate with each other and with computers.
What is an Artificial Intelligence after all ?
Let’s get some concepts down to simplify what is AI and human-computer communication. This is a follow up to some thoughts I wrote down about data formats, Artificial Intelligence, and Algotrithms in 2017.
A basic human-computer interaction is one where we give it instructions or save information in it. Later we can retrieve it or process it. This is a spreadsheet, a word file, and any software that lets us organise work in any way we see fit.
When we give it instructions as formulas or scripts we are building an algorithm, a list of steps to solve a problem or process data. The result of the algorithm together with the context of the data is information, because we are able to make decisions based on it.
Advanced algorithms have been put to practice by Google and Facebook to deliver information and content that we find useful or entertaining. But we forget that even an excel spreadsheet can be an algorithm in this sense.
So what is an Artificial Intelligence ?
We can discuss AI as a computer able to process information as a human, able to communicate in natural language, capable of displaying a reasoning similar to ours. Russell & Norvig, (2020) take this definition step by step and define an Intelligent agent. This is what we are building, agents and assistants that simplify complex and tiresome labour.
From my communication perspective, an AI Intelligent Agent is a system that is able to gather data, turn it into information, and communicate to humans. With the current advances of GPT3, Stable Diffusion, DALL-E, and other models, my definition falls short.
These systems don’t just produce information, they generate knowledge.
If you ask Chat-GPT3 what is a sonnet, the system can explain the concept and apply to write a new piece of poetry.
Bruno: can you write a sonnet about love death and robots?
ChatOpenAI: Sure, here is a sonnet I came up with on the theme of love, death, and robots:
Love, death, and robots, three concepts intertwined,
A deadly trio that we can’t escape,
Love, a force that keeps us close and intertwined,
Death, a shadow that follows us, a shape
Robots, a creation of man’s own design,
But what happens when they love, and what happens when they die,
Do they feel the same emotions as we do, or are they simply lines
Of code, programmed to simulate, to try
To understand the world through their artificial eyes,
Do they love, do they mourn, do they feel alive,
Or are they just machines, without heart or soul,
Love, death, and robots, a puzzle to solve,
But one thing is certain, they will always be a part
Of the human experience, in one way or another,
Love, death, and robots, forever entwined.
There are a number of methods to extract knowledge from a pile of data and information.
- Natural Language Processing (NLP)
- Machine Learning (ML)
- Neural Networks
- Decision trees
- Evolutionary algorithms
- Rule-based systems
- Fuzzy logic
- Deep learning
These are things I only know by name, and for some I have a slight understanding of how they work. For example, Machine Learning is used on Gregory-MS, one of my projects. The Machine Learning algorithm is trainned by me to identify journal articles focused on patient outcomes in the field of Multiple Sclerosis.
As far as I can understand, machine learning algorithms plot the training data on a chart and use probability to determine what label applies to new data it receives.
Natural Language Processing (NLP) on the other hand can break down a sentence and identify the parts of speech. Nouns, verbs, entities, and even biomedical entities if we use the correct ontology. With NLP we are able to give computers a clear language request and get back the equivalent Python code, for example.
I work with databases often and, instead of writting the complex queries, I use Chat-GPT3. And it can go further, AI systems are able to write full blog posts and press releases.
This is where it gets tricky for Communication Professionals. Copywriters, Journalists, Press Agents, Editors, and writers of any kind. What are these professions bringing that is valuable to the process? Their role can’t just be to feed information to the AI and edit the output.
Right now, this is enough for us to have a common basis for the conversation that comes next. Part 2 will be coming soon.
Comments on the blog have been broken for a while, but you can always comment on Mastodon by tagging me. https://mastodon.social/@brunoamaral
References
Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th edition). Pearson.
