They grow up so fast. Especially when they're trained on billions of parameters of data and are used by over 100-million people worldwide.
ChatGPT is officially 1 year old today ?? . So a happy birthday to it and to many more to come. Many of us remember the surprise and shock when it was first announced back in 2022 as we skeptically asked it to multiple 2 numbers together in an attempt to say hello.
How does ChatGPT feel about turning 1 already? I think it's enjoying the process so far.
With that, let's take a look at just how much has gone down at OpenAI in regards to ChatGPT in the past year, because it's alot. The frequent updates, fixes and feature releases have been consistent and enough to keep the platform in the collective zeitgeist.
Update to GPT-4
When ChatGPT first launched, it was using the still impressive GPT-3.5 model, which many people still rely on today as it is faster and has no usage limits so far on the platform.
The release of GPT-4 was a huge step up when it came to understanding context, prompt sizes and in general being able to generate more accurate answers, though be it at a slight cost.
GPT-4 is still only just available to Plus subscribers however, and at the time of this writing, you still have to join a waitlist in order to upgrade. This was mainly due to overwhelming demand occurring after OpenAI's first developer conference a while back.
And more recently, OpenAI announced that ChatGPT would be upgraded to use GPT-4 Turbo, which is more cost effective for developers and came with an updated knowledge cutoff date.
And the rumors of GPT-5 are already rampant and circulating the web as well.
Ability to surf the web
We can't forget the panic and world-ending rumors that circulated when ChatGPT gained the ability to crawl the internet. Much like millions of web crawlers do on a daily basis without anyone being the wiser.
And then, for a short time, that feature was gone probably due to concerns over privacy and other ethical guidelines.
After some fine-tuning the feature is indeed back and while a nice addition to have, I personally have not found it to be too useful in my day to day work.
One of the more exciting features to be released was the ability to generate images directly in the chat prompt allowing you further fine-tune your descriptions and generate more accurate variations.
While there were already a myriad of options for image generation, such as Midjourney and OpenAI's interface as well, none of them were really straightforward to use and some required a bit of technical knowledge in order to get up and running.
But Dall-E on ChatGPT works right out of the box without any extra configuration, training or anything of the sort. It's not the most accurate true to life image generator out on the market today, but it's definitely decent enough for creatives looking to get some inspiration.
Sometime around May of 2023, OpenAI released the official ChatGPT mobile app, first landing on iOS and soon after on Android devices.
The functionality was pretty much identical to the desktop version, though now you could take out your phone and show your friends and family the power of A.I. Which I personally did. All the time.
Slowly but sure though, OpenAI released new features to only select users before making them generally available, which was a brilliant strategy as FOMO is real. And the more people heard about voice capabilities and image generation, the more they wanted to download the app.
One of the biggest updates to land on the mobile app (and on the desktop version) was the ability to recognize images either through files uploaded or through pictures taken from the camera directly.
The use cases, of course, were vast and really up to each individual to figure out. I personally tried everything from reading nutrition fact labels to taking pictures of family and turning them into stickers.
I've also read about people taking pictures of websites and generating the HTML and CSS to render those pages, but for now I'm a bit skeptical about the overall quality of that conversion.
As more integrations come online though, I can definitely see a world where I can take a picture of a product and ChatGPT finds/adds the item to my Amazon cart.
And in the world of accessibility, particularly for the vision impaired individuals, ChatGPT vision could be a game changer in terms of item recognition, such as labels on medicine bottles and even in reading content while on the go.
And lastly, ChatGPT got its voice finally. While initially this feature was only available to paid members, either on the Plus subscription or Enterprise, on Nov 21st, the feature rolled out for every user with the app.
And having used this feature extensively for the past few months now, I can say that it's definitely a whole different experience than simply typing a prompt and getting a paragraph back.
For one, ChatGPT takes on a more human tone when in voice mode. It provides shorter answers, giving more room for a conversation. And it asks alot of questions.
Each time you provide an answer to those questions, you get back a very neutral (somewhat positive?) response ,along with a follow up question. But I'm interrupted its line of questioning many a time and interjected my own, and am constantly surprised at how it navigates the conversation.
And much more
ChatGPT is a great interface for pretty much anyone and everyone to get acquainted with large language models and A.I. in general. And in a year, it's managed to put itself in a very unique place as not only a multi-modal A.I. tool, but one that people haven't given up on yet.
Walter Guevara is a software engineer, startup founder and currently teaches programming for a coding bootcamp. He is currently building things that don't yet exist.
Stay up to date
Sign up for my FREE newsletter. Get informed of the latest happenings in the programming world.