Hello curious people,
Let’s start this with a riddle:
I am a creation of humans,
But I have a mind of my own.
I can learn and adapt,
And I am always growing.
I am not alive, but I am not dead.
I am a tool, but I am also a threat.
What am I?
The answer to this riddle is Artificial Intelligence, and the riddle was generated with chatGPT.
You have probably heard a lot about this tool by now. But today we are going to have a deeper look at it. This revolutionary tool gained over 1 million users in less than a week since its launch. Beating tech giants like Netflix, Instagram, and Facebook in the time it took to achieve this milestone.
Some say it is the next big thing. Something in line with inventions like the printing press, electricity, or the internet. While others are not that impressed.
But what I think all of us could agree on - it is a good sample of where the technology of the future will lead us.
Today we are going to look at:
What is chatGPT
The technology behind it
How we can use it
It’s shortcomings
Pricing
Alternatives
Let’s begin with a short introduction.
What is chatGPT
Here’s chatGPT itself answering the question.
And on a surface level that’s about it. But chatGPT is capable of a lot more than simply generating human-like texts. It can summarize complex texts, explain things in an easy-to-understand manner, it can compose e-mails, come up with schedules and even generate code that actually works. We will explore all of these use cases and more further down in the article.
But it also has its shortcomings. And even the CEO of OpenAI, the company behind all of this, acknowledges that chatGPT is lacking in some areas.
So before we go any further I would like to remind all of you that chatGPT is not perfect. And its outputs shouldn’t be considered factual. Although sometimes it’s right, sometimes it can and it will spit out complete nonsense.
The Technology
ChatGPT is based on one the largest LLMs (Large Language Models) out there - GPT3.5, it’s the latest release of GPT currently available to the public. The model is trained using RLHF (Reinforcement Learning from Human Feedback).
If you are looking to learn about all of this in more detail, you can visit OpenAI’s blog post about chatGPT. And get the info straight from the source. I will try to keep things simple here.
The main difference between chatGPT and previously available versions of the GPT language model is its ability to answer follow-up questions. You can keep a conversation going and chatGPT will operate within the whole context.
While technology like this is really impressive, there still are some major limitations. We will look more into those after we are done with in my opinion the most exciting part. The potential ways we can use this technology.
Capabilities and Use cases
By now the internet is full of examples of people using chatGPT in various creative ways. Today I will share with you some of my favorites.
These range from simple tasks, like answering questions to something like passing the Bar exam or even building an app from scratch.
Let’s check out some examples
Here’s someone using chatGPT to summarize Netflix’s earnings report
This person tested if chatGPT could potentially pass the Bar Exam
And here’s a video on Youtube of someone creating a To-do list app from scratch with the help of chatGPT
I asked chatGPT to act as a python tutor and also simulate the console when I type in code. Here’s a short example of that:
The possibilities feel endless. You can ask for advice on just about anything. And then dig deeper with follow-up questions. People use this to come up with business ideas, manage their schedules, improve their diets, debug their code, and much more.
I could go on with examples for a while. But I think this will be enough for today. I have compiled some more creative use cases in one of my Medium articles. And it’s still just scratching the surface.
Some Tricks and Tips
A helpful list of ready to use prompts - link here
Use follow up questions - this will increase your productivity using chatGPT
Point out the mistakes in chatGPT outputs. - Respond to it with what you didn’t like, in some cases this will solve your problem instantly
Ask chatGPT to elaborate on specific parts of it’s outputs to get more detailed information
Limitations
While chatGPT definitely is impressive. It’s not magic. It’s just a Large Language Model. It is not a sentient being. And it sometimes outputs nonsense. Here’s one great example of it’s shortcomings.
At first glance it might seem that chatGPT can do just about anything, but once you have used it for a while, you begin to see more cases similar to the one I showed you above.
Don’t get me wrong it still has a huge potential for boosting productivity. But it does have it’s limitations and shortcomings. Here’s a list of things I feel it’s lacking.
Ability to check the reliability of the facts it presents.
Access to the internet and up-to-date information
Ability to save conversations
And I do believe all of these could be solved over time.
Pricing
We have been given a chance to use this emerging technology for free. But I don’t think this is going to last forever. And we will probably see a pricing model similar to that of GPT-3 in the future. It might be more expensive because of the added features.
Currently, the state of the art Davinci GPT-3 models are priced at 0.02$ per 1000 tokens. 1000 tokens get’s you about 750 words. So the amount you would have to spend would vary a lot depending on how you use the tool.
Alternatives
Currently there are no direct competitors to chatGPT available to the general public. But there’s definitely competition. OpenAI’s advantage is that they don’t need to worry as much about their reputation as established tech giants like Google.
This is one of the main reasons stated by Google parent company’s CEO Sundar Pichai, as to why they won’t be releasing similar products to the general public anytime soon. But it is no secret, that Google has a pretty impressive LLM of their own. And they are working hard to get this technology consumer ready and fail proof.
And there are also some opensource alternatives to GPT-3, like GPT-J or GPT-Neo. But these are still nowhere near as good as GPT-3, and they don’t offer as user friendly experience as chatGPT does.
Closing Thoughts
In our daily lives, there are loads of tedious, time-consuming tasks most of us despise. Some of these tasks involve generating text, or code. And AI can make it easier for us. This technology is here to stay, and it will only improve over time.
Tools like chatGPT make it a lot easier to handle complex tasks like writing code. And the trend in general for AI based tools is to lower the barrier of entry to various fields of expertise. Here’s the perspective of openAI’s CEO Sam Altman:
While we are not quite there yet. We have been given a glimpse into the future with chatGPT. And we should definitely use that. We need to adapt, to stay relevant in this everchanging environment.
Thank you for reading!
Stay curious and stay HungryMinded!