The progress in the artificial intelligence sector has been increasing tenfold. Amidst the changing landscapes, OpenAI, founded by Sam Altman, has now introduced a significant upgrade to ChatGPT‘s image generation capabilities. This change, which is available to users from today, is not just alarming but will likely further affect image-makers around the world. Here’s a look at what the new update can do.
According to Tech Crunch, Altman announced the update during a live stream on Tuesday. Per reports, ChatGPT can now use the GPT-4o model to create images and modify them natively. This means the software can help you make realistic images within the chat. GPT-4o was able to create and edit texts alone before this, with the new image generation seemingly a way to beat xAI, a model created by OpenAI’s Chinese competitors.
Furthermore, this native image generation will be available on ChatGPT and Sora, the latter being for video creation, for a $ 200-month subscription to the Pro plan. The feature will gradually be released to Plus and free users of the AI software soon, as well as its API services.
What makes this update so challenging is that the new update creates more accurate and detailed images. In addition, the GPT-4o model can edit existing images, whether products or people and transform them into something else. One can also simply edit details in the foreground and background.
In another report by the Wall Street Journal (WSJ), OpenAI trained the new model on “publically available data” and data from Shutterstock, their partner in this matter. However, this also means that this kind of data training can also lead to IP-related lawsuits, and so many companies do not reveal how they achieve such a feat. Instead, OpenAI’s chief operating officer Brad Lightcap told WSJ: “We’re respecting of the artists’ rights in terms of how we do the output, and we have policies in place that prevent us from generating images that directly mimic any living artists’ work.”
However, the most interesting part is that one has to inform OpenAI that they wish to “opt-out” of using their work to train its AI. How so? By submitting a form! However, this form has been changed as well as of October 2023, and instead, one has to use a button. As ridiculous as it sounds, it is quite possible that your work still can be scrapped, including web-scrapping, which means bots collect data, including images, from one’s websites.
Furthermore, with Google’s Gemini 2.0 easily removing watermarks, the future of photography seems to be more in trouble. The lack of guardrails, consent, and credit and copyright challenges further make the advancement of AI a threat to the livelihoods of photographers. Furthermore, it makes it increasingly easy to recreate the works of other photographers without giving them credit or compensation. The picture of the Afghan girl is an example of that. This also proves that OpenAI is scrapping the internet, as it clearly trained its AI on McCurry’s image.
Prior to this announcement, many tech companies were also lobbying for more lenient rules for AI, including access to copyrighted materials to train their AI, reports The New York Times. This means that everyone, including Google, OpenAI, and Meta, is aiming to get access to restricted material, thanks to Donald Trump’s announcement that AI will be used to beat China. There is also a push to use publically available data as “unequivocally fair use.”
Considering the current scenario, use Nightshade and another such model for your images to stop AI from scrapping your wor. You mus also actively watermark and search the web if your picture has not been used by bad agents. It is an uphill battle, but every small decision you take can impact your photography. Until then, be vigilant and keep an eye out for more changes.