Discover what AI really is and the benefits that it can provide us, solopreneurs.
Want to discover 10 AI tools that will give you an edge as a solopreneur?
Visit https://macpreneur.com/ai to grab a PDF containing the TOP 10 AI tools every Macpreneur should know about.
Artificial intelligence (AI) is the intelligence displayed by machines or software. It is a field of study in computer science that focuses on developing and studying intelligent machines.
AI technology uses neural network modelling to mimic the functioning of the human brain. AI tools are designed to make predictions based on training data by recognising patterns and features.
Building an AI tool involves four phases: deciding the task, designing the neural network, training the model, and validating the output. AI tools can be single-modal or multimodal, trained on generic or domain-specific datasets.
They are not intelligent or creative themselves but enable human creativity. AI tools can save time and money, repurpose content, and enhance productivity.
AI technology is continuously evolving and will have a profound impact on society.
- Artificial intelligence (AI) refers to the intelligence displayed by machines or software.
- AI is a field of study in computer science that focuses on developing and studying intelligent machines.
- AI technology uses neural network modelling to mimic the functioning of the human brain.
- AI tools make predictions based on training data by recognising patterns and features.
- Building an AI tool involves deciding the task, designing the neural network, training the model, and validating the output.
- AI tools can be single-modal or multimodal, trained on generic or domain-specific datasets.
- AI tools are not intelligent or creative themselves but enable human creativity.
- AI technology is continuously evolving and will have a profound impact on society.
It’s not a robot, it’s not even intelligence, And it’s certainly not a threat. So what exactly is A.I.?
I’ll unpack all of this after the intro.
If this is the first episode that you’re listening to, welcome to the Macpreneur Tribe, and if you are a longtime Macpreneur listener, thank you for tuning back in.
As a fellow solopreneur, I appreciate that you dedicate these 15-ish minutes with me every week.
So let’s start by defining AI.
And here is the first paragraph of the Wikipedia article on artificial intelligence.
So I quote, “Artificial intelligence (AI) is the intelligence of machines or software as opposed to the intelligence of humans or animals. It’s also the field of study in computer science that develops and studies intelligent machines. AI may also refer to the machines themselves.”
But who coined the term for the first time and when?
The origin of A.I.
And so in September 1955, a group of mathematicians and computer scientists submitted a funding proposal in which they coined the term artificial intelligence.
The funding was accepted and led to a research project on artificial intelligence that took place during the summer of 1956.
And so this thing started about 70 years ago, and it took quite a long time for it to progress because at the beginning, they didn’t have the algorithms and the compute power that we have today.
The technology behind A.I.
And so let’s talk about technology.
AI tools leverage a technology called neural network modeling, which is a fancy way to say, let’s mimic the way that the human brain works using computers.
And in case you didn’t know, before being a solopreneur, I was working as a research and development engineer for one of the biggest tire manufacturers in the world.
And I used neural network modeling to help design tires that should meet specific performance characteristics.
And so in its simplest form, a neural network is composed of multiple layers of digital neurons.
So there is an input layer, then a series of hidden layers, and then an output layer.
You might have heard the word deep learning.
It refers to neural networks that are made of dozens and sometimes more than a hundred hidden layers.
What an A.I. tool does
And so, what’s the goal of an AI tool?
It’s simple. It is to make predictions.
So, like the human brain, it can only make predictions about something that it has been trained on before.
For example, if you have never learned Chinese, you won’t understand any word that you hear or read.
But after learning Chinese, your brain becomes capable of converting spoken or written Chinese words into an English word.
Another example is recognizing animals.
Your brain has seen enough cats and dogs that now you can recognize which is which, even if it’s an animal that you’ve never seen before.
And it’s not because your brain has a database of pictures of cats and dogs that it goes through and quickly tries to compare what you see.
No, it’s because the way the brain works is through pattern and feature recognition.
And so the brain is capable of recognizing edges, angles, distances, and shapes in such a way that if I show you a picture with a cat and a dog, even though both animals have fur, whiskers, four legs, and a tail, your brain has learned to recognize the differences in the distance between the eyes, the position and shape of the nose, the shape of the ear, and the position of the ears on the head.
And so, for Artificial Intelligence tools available today, it’s exactly the same.
AI can’t think. It’s not aware of its own existence. It’s just a computer program that has been trained to predict an output based on an input.
And so the next question is, how are those AI tools built or created?
How A.I. tools are built
Creating or building an AI tool is done in four phases.
Phase number one is to decide the type of task that the tool will perform.
It could be either pure processing or it could be generation, and we are talking about different kinds of medium.
So for instance, processing text can help detect the mood or the sentiment inside a conversation.
For image processing, it could be to detect faces, find objects, or recognize animals inside a photo.
For the generation part, it’s an AI tool that would be able to generate text, images, videos, music, speech, and this is where the term Generative AI comes in.
Once the task has been decided, phase 2 is to design or choose the size and configuration of the neural network.
In a simplified version, it’s just how many neurons will the network have, and how many layers will there be between the input layer and the output layer.
And once the model has been decided and configured, phase number three will be training the model.
And here again, there are two options. We can train a model on data that has been labeled by a human.
So for instance, we could feed and train an AI model based on images, a bank of images that contain animals, for instance, and every image has already been pre-labeled.
So the type of animal, the breed, maybe some characteristics, the color, and stuff like that.
But the other option is to train the model on what we call unlabeled data.
It could be, for instance, a large corpus of text.
And so, in this case, this is what we call pre-training.
And afterwards, a human or a group of humans would do some fine-tuning.
Now, ChatGPT, you might be wondering what GPT stands for.
Well, it’s Generative Pre-trained Transformer. And so the P in GPT is that.
So, ChatGPT has been pre-trained on a bunch of text and afterwards, there has been a fine-tuning phase.
And then the last phase, phase number four, is the validation phase, where you give the AI tool something that it has never seen before, so it was not part of the training, and you evaluate the output.
And going through these four phases, and then deciding to go back to phase number one or phase number two, it can take months and a lot of compute power, which is the reason why OpenAI took five years to iterate between GPT-1 that was released in 2018 and GPT-4 that was released in 2023.
The modality of A.I. tools
One way to think and look at AI technologies is to consider the number of tasks that a neural network can perform.
So a lot of them can only perform a single task.
So version 3.5 of ChatGPT can only interpret and generate text.
StableDiffusion and DALL-E are image generation tools. They can only do that.
But now, more and more, AI tools and applications are what we call multimodal.
In other words, they can do more than one task.
At the time of recording, ChatGPT version 4, which requires a ChatGPT Plus subscription, can interpret text, images, and voice, and as the output, it can obviously generate text.
But now it can speak out loud and it can even generate images because it has direct integration with DALL-E, which is the image generation tool offered and created by OpenAI as well.
The domain(s) of A.I. tools
Another perspective that can help understand the world of AI technologies is to consider the domain or the field in which the neural network has been trained.
So some AI tools, and I would say most AI tools, are trained on vast and generic amounts of data, which means that they can be useful in many industries, which is the case for ChatGPT and Google Bard.
Both of them belong to a class of models that we can refer to also as an LLM, for a large language model.
Now even though we are dealing with a more generic AI tool, it’s possible to narrow the domain or condition it to be more precise by prompting it in a specific way.
So for instance, if you need to create a Facebook ad campaign, you could immediately ask ChatGPT to produce some copy for you.
But the best way to have the best copy is to start by telling ChatGPT, “you are now acting as an online marketing expert specialized in Facebook ads.”
Just adding this short sentence at the beginning of your prompt will give you much better results because you’re conditioning ChatGPT to produce an output that has been narrowed down to this specific expertise.
But there are other AI tools that are trained on a narrower and more targeted dataset, and thus are referred to as being domain-specific.
So for instance, Bloomberg released something called Bloomberg GPT, which is similar to ChatGPT, but it was trained mostly on financial information and financial news.
And so now let’s talk about what AI is not.
A.I. tools are not intelligent
AI is not intelligent.
In the case of generative AI tools like ChatGPT and DALL-E for images, the output has been designed to look like what a human would write or draw or paint, which is why most people have the impression that they interact with something that is intelligent when in fact, we’re just interacting with a highly sophisticated mathematical model.
A.I. tools are not creative
And therefore, AI is not creative, but it’s rather an enabler of creativity.
Back in June this year, 2023, I attended a conference in Luxembourg, and there, one of the speakers, Dave Coplin, said something quite profound.
He said, “AI will take the robot out of the human.”
And what he meant was that AI can help us automate repetitive cognitive tasks that require very little creativity.
And so, as a result, we, humans, will have more cognitive bandwidth for the tasks that actually require creativity.
Another way that AI can help us be more creative is by offering multiple points of view or different angles when we are confronted with a problem.
It’s something that I’ve been doing more and more now, is to give the same prompt to ChatGPT version 3.5, ChatGPT version 4, and Google Bard.
Because then I get more than one output, and then I can decide and look and contrast and compare and maybe mix and match.
And so it’s a little bit like brainstorming with three different people rather than one person or one machine.
A.I. tools are not robots
So, talking about robots, AI isn’t a robot either because by definition, a robot is a machine that can exert forces and movement.
I find it quite amusing when I see the image of a pensive humanoid robot at the top of an article about artificial intelligence.
Because a more realistic representation would be a computer room full of servers somewhere in a data center.
Having said that, there are already robots that are equipped with AI technologies in order to understand their surroundings and to interact verbally with humans.
And so there is no doubt in my mind that combining robotics and artificial intelligence is inevitable.
A.I. tools are not a threat
The last thing that AI is not is a threat, in itself.
It’s just a tool, right? It’s like a digital hammer.
When you need to hang a painting on the wall, the fastest and easiest way is to use a hammer to put a nail into the wall.
But someone else can take a hammer and throw it into a window to break the window.
The hammer didn’t do anything, right?
It’s the human who is misusing the hammer and then becomes a threat.
And so with AI technologies, it’s the same.
It can be used for good things or it could be used for bad things.
Now, some of you might be thinking, “But Damien, because of AI, surely my job is or will be under threat.”
If, like me, your biggest asset is your brain and your knowledge, then it really depends where that knowledge is coming from.
If it’s coming from experiential learning, in other words, if your knowledge is based on years of experience, there is very little chance that any human, without the same knowledge as you, even equipped with the best AI tool out there, can compete with you.
On the other hand, if the knowledge that
you have accumulated is actually readily available in any way, shape, or form on the internet, then, indeed, anyone who has access to an AI tool that has been trained on the same information that you use right now will indeed become tough competition to you.
For example, as an Apple expert, I used ChatGPT to help me solve a very specific problem that was quickly filtering videos from the Photos app based on their size.
Now I knew from experience that the Photos app does not offer this capability, has never offered this capability natively, but ChatGPT still managed to craft a step-by-step procedure involving menu items that don’t exist in the Photos app.
And when I asked it, “Look, it does not exist.” ChatGPT told me that I must have the wrong version of Photos or the wrong version of macOS, which obviously was completely ludicrous.
The actual solution required reusing and adapting an Apple script that somebody else wrote a few years ago for a different reason.
But that script had the necessary ingredients nonetheless.
And so, for this kind of problem, only another Apple expert with the same experience as I had and coding skills as mine could really compete with me.
So rather than looking at AI as a threat, I prefer to see it as an opportunity with many benefits.
The benefits of A.I. tools
The first benefit is that it’s a time-saver.
Now you can use ChatGPT to quickly outline a blog post or a newsletter article, and it can even provide you with a first draft.
Once you have fleshed out the article, you can use ChatGPT again to quickly summarize the article, extract key takeaways, and even prepare social media posts with emojis.
For example, before ChatGPT was a thing, it took me about five hours to craft the EasyTECH newsletter in two languages, English and French.
It was a bit too much, so I stopped manually creating the French version and instead used Google Translate.
I was not even using Google Translate to recreate a French version.
I was just providing the French-speaking subscribers with a URL that was Google Translate blah blah blah slash the URL of the English version of the articles.
I was not doing any work at all except providing that URL.
Doing that helped me reduce the effort to 3 hours and 40 minutes, saving me about 25 percent of the time.
But the Google Translate version was not super good.
After a while, I decided that I still needed a French version, and ChatGPT came along.
So I started using it for the English version.
To flesh out the first draft of the individual topics, which saved me a lot of time, and then I asked ChatGPT to translate from English to French, and I was amazed by the quality.
It was not perfect, so I had to read everything and then tweak a few sentences here and there, but it was much better than Google Translate.
And now, to produce the same output as before, the newsletter in two languages, it only takes me roughly three hours.
So it’s a 40 percent time saving for the same output as before.
Another benefit is that it can save money.
Transcribing a podcast used to cost a lot of money, either through hiring a VA or using a paid service like Rev.com.
So before AI transcription became mainstream, Rev.com was charging $1.25 per minute of audio, which would then be roughly $25 per a 20-minute episode.
Now, with a tool like MacWhisper, you can transcribe audio and videos for free directly on your Mac.
And best of all, you don’t need to upload anything to the cloud. So it’s 100 percent private.
The third and last benefit is that it can help you quickly and easily repurpose your content, allowing you to publish more in less time.
If you’re following me on Instagram, Facebook, LinkedIn, or TikTok, you have seen short-form vertical videos with captions and everything.
All those are extracted from the video version of the episode.
At the beginning, I was using Descript to prepare that, but it was still a lot of manual process and manual work.
Now I’m using a tool called Opus Clip, which takes the YouTube link of the episode, then analyzes the video, and isolates the most interesting snippets.
So it cuts the video into smaller segments, and for each segment, it gives me a virality score and an explanation of why it believes that snippet will grab the attention of the viewers.
What is nice with Opus Clip is that I can specify the length of the clip.
So I actually do two passes.
The first one is 30 seconds to 60 seconds for Reels and Shorts.
The second time I use Opus Clip is with a length of 90 seconds to two minutes.
And now I can reuse those on TikTok and LinkedIn.
On top of that, I have the ability to tweak them, so I can review the transcription and correct it if needed before downloading the clips.
Even though it costs a little bit of money, the base plan is $19 per month for three hours and 20 minutes of video, it’s still much cheaper than hiring a video editor.
And I get social media clips for multiple platforms in one go.
So to recap, artificial intelligence can be many different things, but at the heart of it, it’s a field of study that leverages machine learning and neural networks to create computer programs that mimic the human brain and thus appear to be intelligent.
The primary purpose of AI tools is to make predictions that can be used either to process an input or to generate something new.
AI technology can be used for various media types: text, images, sound, video, computer code, and the landscape of AI applications is continuously evolving.
We’re really in the early stages of AI adoption, and I firmly believe that AI will have as profound an impact on our civilization as electricity had when it came around.
Sure, some jobs will disappear, like the people who used to light lampposts at night with oil lamps, but also a lot of new jobs will be created, and in organizations, small and large, there will be people creating and training AI models on proprietary information.
In the near future, every aspect of our life will be enhanced by AI somehow, and it won’t be advertised as such anymore.
The same way that when you buy a washing machine today, it doesn’t advertise the fact that it’s powered by electricity.
In the meantime, solopreneurs who start leveraging AI tools will save time and money almost immediately.
Now, if you’ve never used any AI tool before, I hope this episode has motivated you to give it a go.
If you have a Google account, just visit bard.google.com.
Otherwise, I suggest creating a free ChatGPT account via chat.openai.com.
And if you’re already leveraging AI in your solo business, please send me a DM on social. I’m @macpreneurfm on the main platforms, and let me know which AI tool you’re using and how it’s helping you streamline your business.
So that’s it for today.
In the next episode, Episode 69, I will dive deeper into ways that AI can help solopreneurs save time and money.
In the meantime, I have prepared something that you may find useful.
It’s a PDF that contains the top 10 AI tools that I believe every Macpreneur should know about.
You can grab a copy at macpreneur.com/ai and learn all about these top 10 AI tools, or you can click the link in the show notes.
Until next time, I’m Damien Schreurs, wishing you a great day!