As artificial intelligence-assisted vehicles gain new skills every day, the idea that people will “take our jobs” is getting stronger in direct proportion. But interestingly (and as it should be) lately, this situation has started to turn into an effort to adapt, rather than fear. In this context, artificial intelligence tools are no longer perceived as a threat, but as an assistant.
It would not be wrong to say that GPT-4, which was recently announced by OpenAI, has a dynamic quality for this very reason. You have often seen the previous model of this technology, which we will go into details. The answers he gave to the questions, the sites he created, the applications and many more came up frequently. So what changed in the new one?
GPT-3 contained 17GB of data inside.
For GPT-4, the size of this data used in training is around 45 GB. This shows that it will have a wider response capacity as well as being consistent in answers. Apart from that, if we come to the parameter issue; GPT-3 contained 175 billion parameters. Isn’t it too much? It doesn’t sound like that much when you hear the 100 trillion parameters of GPT-4.
- 175 billion = 175,000,000,000
- 100 trillion = 100,000,000,000,000
It is claimed on the internet that this huge parameter difference will bring the new model closer to the human brain, but it will not be very consistent to give it in a precise language unless we experience it first hand. However, you can be sure that you will get more consistent, natural and comprehensive answers compared to the answers you get from ChatGPT.
Apart from that, this model can read, analyze and write up to 25,000 words.
In addition, OpenAI CEO Sam Altman stated that GPT-4 will not exceed GPT-3 in terms of size.
In addition, there is no doubt that the new model will be faster than its predecessor, as it will work with much more powerful CPUs and GPUs.
Finally, let’s clear up some confusion before we move on to real-life scenarios. GPT-4 is a model that you can only process with text. For this reason, it cannot be used in visual or video production like DALL-E. So you can create your video text, but you cannot create your video.
But that doesn’t mean it stays put when you give it an image.
Because this model also has the ability to understand by analyzing the image. In other words, when you give him the image of two cars coming towards each other, he can indicate that the cars will collide. Or when he sees a plate that is about to fall in a photograph, he can say that this plate will break. Or, he can understand exactly what is funny in a funny image.
In the image above, “What will happen when the glove falls off?” He answers the question that the ball will take off.
Everything is fine on paper. But the difference in real-life scenarios is the most important part for us.
In the live broadcast where the new model was announced, the site description on a ‘handkerchief’ had become reality.
Desired items such as “(Here’s a funny joke)” are placed in certain places and this image is transferred to the GPT-4 model as a description. In a very short time, the codes of a fully functional website suitable for this description are given.
Even the simplest games can take hours to make. However, a user obtained a copy of the iconic Pong game in 60 seconds with their codes.
Another user recreated another iconic game, the snake game.
A user with no coding knowledge created a plugin for Google Chrome with the help of GPT-4.
Duolingo has announced the GPT-4-based subscription level called Duolingo Max.
There are two main features that stand out in this subscription. The first of these, ‘Explain My Answer’, explains to you why the wrong answers you gave were wrong. In this context, you can avoid minor question marks and repetitive mistakes that remain in your mind.
The second one comes in the name of role-playing. Earlier, we said that the claims that GPT-4 approaches the human brain are back. By taking advantage of this, you have the opportunity to make real-life conversations with artificial intelligence, which makes it a much more productive practice than usual.
According to Dan Shipper, he can create new derivatives from an existing drug.
For example, it can take the components of an existing drug and give other components with the same properties. He can use different ingredients to avoid patenting, and he can even order drugs online – even by sending a purchase e-mail to the seller.
A user named Jackson Greathouse Fall doubled his $100 in a short time with the help of GPT-4.
Saying that he has 100 dollars for the artificial intelligence model and that he will increase it legally as soon as possible, Jackson spends on the answers he receives from GPT-4. Receiving a recommendation to open a website in the field of sustainability, Jackson has this website built by artificial intelligence. Even the brand logo comes out of artificial intelligence. GPT-4 sets the advertising costs again. In other words, the management of the $ 100 is completely in the hands of GPT-4.
The articles on the site are also prepared using real sources through artificial intelligence. By the end of the job, Jackson states that he currently holds more than $1300 in cash. But it’s worth noting that Jackson did this experiment openly. In other words, this money, which came with the contribution of investors, was perhaps made by people who were aware of the event, out of curiosity, we do not know. Still, it’s a good example that showcases the power of GPT-4.
It has been announced that the next model that Bing will use in its artificial intelligence assistant will be the GPT-4.
This means that we will be faced with a smarter assistant. We will see together whether it will come to the fore as before.
Intercom company announced its new product based on GPT-4: Finnish
It will appear as a GPT-4-based client representative.
Supporting the visually impaired or visually impaired, Be My Eyes announced its new assistant called Virtual Volunteer™:
As can be seen from the video, the assistant in question can specify the details of the clothes, understand the plant type, and guide the user, making it possible to navigate around the room or a building without getting lost. In addition, after responding, if more details are needed, he can easily provide it.
As you can see, GPT-4 has already opened countless doors.
After that, we will see what comes next and we will pass it on to you. Soon, our original content will also be on the site. You can share your thoughts on this subject and the content you want to see in the comments section.