OpenAI's 4o Image Generator, the Next Big Thing, or Another Flop?

AI Riddle:

I can exist in a second, stretch for years, or vanish in a moment. I am measured but never seen, spent, and can be saved.

What am I?

Today’s Motive:

🖼️ OpenAI’s next-level image generator is out to the public for free. But, is it better than the competition?

💻️ AI crawlers are breaking the internet. Is there a way to stop this?

🧠 UC Berkeley Engineering implants an AI device to help non-speakers communicate.

🤑 Check out the latest AI deals below. 💸 

🛠️ Get the scoop on the latest AI tools in the at-a-glance section.

OpenAI

🚀 Are OpenAI's New Image Enhancements As Groundbreaking as Everyone Says?

Let's start with the image above, created with Midjourney. OpenAI's image generator couldn't give me anything near what I wanted after several attempts, and then stopped working.

 🙃 OpenAI has recently enhanced ChatGPT's image generation capabilities by integrating the GPT-4o model, a natively multimodal system capable of processing text, images, and audio.
For further reading, follow the link here.

The Direction

This integration allows ChatGPT to generate high-quality, photorealistic images directly within the chat interface. ​

Improvements

One notable improvement is GPT-4o's ability to accurately render text within images, a challenging task for previous models like DALL-E. You can now create images containing precise textual elements, enhancing utility. ​

Additionally, GPT-4o supports inpainting, enabling you to edit existing images by adding or removing elements with text prompts. This feature allows seamless image modifications, such as altering backgrounds or inserting new objects, directly through conversational interactions. ​

I tried the watch test to see if OpenAI could break previous barriers by creating something new. Try this test with other AI image generators. Every image generator will show wristwatch hands at around 10:10. Typical images of watches on the internet have the hands around this time.

As you can see, ChatGPT did a poor job of developing a stunning display of product photography, and the time was supposed to be 8:45.

Free Tier Upgrades

These advanced image generation features were available exclusively to ChatGPT Plus and Pro subscribers; however, as of March 31, 2025, OpenAI has extended access to all users, including those on the free tier, with certain usage limitations. ​

The introduction of GPT-4o's image generation has led to a surge in user engagement, particularly with the creation of images in the style of Studio Ghibli.

This trend has significantly increased ChatGPT's user base and subscription revenue, highlighting the impact of these enhancements. ​

The Motive:

Brace yourself for a wave of AI-generated Ghibli-style images flooding the internet—at least until the novelty wears off.

I've been testing the tool over the past few days, and here are some initial thoughts:

Strengths:

It excels at generating text within images, photorealism, the inpainting feature worked out well, and the conversational interface for text-prompts allowed for tweaks and reworks. Although sometimes it felt a little long-winded.

Weaknesses

Image generation is currently slower than other models. I miss the ability to generate multiple images at once. While it can produce photorealistic results, consistency is hit or miss. Overall, other image generators still offer a broader creative range.

While many praise it as revolutionary, I find its guardrails limiting, and its capabilities now feel more on par with the competition.

This follows a familiar pattern in the AI world: a new tool launches, everyone hails it as the next big thing, and once the hype fades, we return to the tools that actually meet our long-term needs.

Today’s top AI tool at a Fraction of the price.

See oncely.com for more deals on AI tools.

AI CONTROVERSY

🕷️ How AI Crawlers are Destroying the Internet

 🕸️ Artificial Intelligence (AI) crawlers, also known as AI bots or scrapers, are automated programs designed to scan and extract data from websites.

While they serve purposes such as training AI models and aggregating information, their rapid proliferation and aggressive data collection methods have raised significant concerns about their impact on the internet's stability and integrity.​
For further reading, check out this link.

Summary:

Overwhelming Web Traffic

AI crawlers have been reported to generate massive amounts of web traffic, sometimes accounting for up to 80% of a website's visits. This surge in automated traffic can overwhelm servers, leading to degraded performance or even causing websites to go offline.
For instance, some websites have experienced nearly a million hits from AI bots in a single day. ​

Evasion of Standard Protocols

Traditionally, websites use a robots.txt file to communicate with web crawlers about which areas should not be accessed. However, many AI companies have been found to ignore these directives, scraping data without permission and exacerbating server strain.

Financial and Operational Burdens

The influx of bot traffic imposes additional costs on businesses, including increased expenses for hosting, bandwidth, and security measures. In 2024 alone, companies reportedly lost approximately $238.7 billion due to bot-driven costs, encompassing hosting fees and wasted advertising expenditures.

Content Appropriation and Ethical Concerns

AI crawlers often harvest content without compensating or crediting the original creators, leading to ethical dilemmas and potential revenue losses for content producers.

Mitigation Efforts

In response to these challenges, some tech companies have introduced tools to help website owners block or manage AI bot traffic. For example, Cloudflare launched a feature that allows you to block AI bots and scrapers with a single click, but how long will these measures last?

The Motive:

The AI attack on your information is a digital warfare going on 24/7. While people like Sam Altman from OpenAI have repeatedly pleaded for absolute access to information to aid AI training, can our systems handle it, and do you want to be another data point in the AI training machine?

I don't think so. AI developers should be trying to figure out the brakes before they stomp on the AI gas.

AI Tools at a glance:

📓 Asana.com: Connect work to goals and automate workflows with AI as your teammate.

🔤 Wordtune.com: Wordtune's Chrome Extension — your all-in-one tool for rewriting, proofreading, and perfecting text, right in your browser.

🗣️ tldv.io: In a meeting? Let AI run and do your hard work.

AI SCIENCE

🧠 AI Neuroprosthesis Restores Natural Speech

 🗣️ New insights from Berkeley Engineering announce a significant advancement in brain-computer interfaces for individuals with severe paralysis who have lost the ability to speak.
Full story from UC Berkeley Engineering, link here.

Summary:

Researchers have developed an AI-powered neuroprosthesis that can decode brain signals related to speech production and synthesize them into naturalistic, intelligible speech in near real time.

This new method overcomes previous limitations by significantly reducing latency, allowing for a more fluent and volitional communication experience.

This highlights the technology's potential to dramatically improve the quality of life for those with speech-affecting paralysis and discusses future directions, such as incorporating emotional expressiveness into the synthesized voice.

The Motive:

This is a great benefit for humanity because it offers a new way for individuals who have lost the ability to speak to communicate effectively. The technology overcomes the previous challenge of latency, allowing for more fluent and natural conversations.

Furthermore, it uses AI to decode brain signals from the motor cortex into speech and can even synthesize words not in its training dataset, indicating a deeper understanding of speech production.