I don't deny that generative AI is affecting the creative sector, as I hate it. Openai's ChatGpt bot, which includes Dall-e's image generator, is one of the major contributors to it. Currently, less than three years after its initial release, another upgrade is available.
GPT-5 introduces a more customizable interface that allows users to choose colours, and even bot personality (presets include “sarcasm”, “geeks” and “listeners”). There is a new “advanced speech” system that is supposed to adapt to the tone, and Openai believes the model itself will provide more powerful inference and improved accuracy and safety. But what does change mean for creatives?
Some users claim they've noticed improvements in image generation, but Openai doesn't say that it made any changes to the image generator with this update, so it's probably running GPT-4. That said, the GPT-5 is said to have better visual perception. This could mean better understanding of reference images.
The main changes come to creatives that use GPT as an assistant for brainstorming ideas, find answers to problems, and as an assistant for generating prototypes for UI and web design. Changing the heading is “built-in.”
The new model will make the problem work step by step before you can get an answer. It basically shows that it works just like you were taught to do it in school. This will help you to see the logic behind the answer and learn what's interesting. It also helps inspire alternative ideas, if you are looking for creative campaign ideas or ways to approach a specific audience, for example.
Openai places a lot of emphasis on the atmosphere coding capabilities of the GPT-5. This is demonstrated by showing that bots can generate apps or simple platform video games from text prompts by specifying features, UI design, art style, and more.
In the video game above, prompt asked for a game about rolling balls with speed, high score tracking, retry buttons, colorful UI, and parallax scrolling backgrounds.
Openai says the new model “can create beautiful, responsive websites, apps and games with an eye on aesthetic sensibility at one prompt.” Early testers like Pietro Schirano (video below) from Magic Path AI show that the results better understand things like intervals, typography, blanks and more.
Interestingly, Openai says that the GPT-5 is more “honest” than its predecessor. The big problems with AI models tend to lie rather than admitting that they don't know the answer. That's because in their training they are rewarded with giving answers. Perhaps GPT-5 will recognize that it cannot complete a task more accurately, communicate its limitations more clearly, and have a lower “deceit notification rate.”
GPT-5 is now available with ChATGPT for users. This includes users with the free tier. There is a usage that requires users to use the GPT-5 Mini. The developers have three versions: GPT-5, GPT-5 MINI and GPT-5 NANO, with different speeds and costs.
For more information about this week's Gen AI News, see Google's Gemini AI Storybook Generator and Disney's AI Setback News.
