Sam Altman wasn’t afraid to try things. That people want to create Studio Ghibli style images? Forward. AI Videos hyperrealistic? Go for it. A browser with AI? we have it. Wherever I saw an option to add AI, OpenAI added it. But that was before, because these projects are being put on the back burner or directly closed for a simple reason: they are blank bullets.
ChatGPT is not going to flirt with you. According to the Financial TimesOpenAI has canceled its plans to launch an erotic chatbot, and now the goal is to focus its resources on its most important products. The decision is partly a response to tensions and internal criticism from employees and investors when offering sexualized AI content. One former employee noted that “AI shouldn’t replace your friends or family; you should have human connections.”
Making an erotic chatbot is not that easy. In addition to the social impact, it seems that OpenAI has had to face really complex technical challenges when creating this type of chatbot. Training an AI model to do something that “normal” models were trying to avoid was causing problems. For example, when including data sets with explicit content it was necessary to eliminate illegal behavior, such as bestiality or incest. That adult mode, called “Citron mode” internally, could have required users to prove that they were over 18 years old.
Too much risk. The move to launch an “adult mode” of ChatGPT was reputationally risky, and people familiar with the decision have indicated that OpenAI wants to begin a long-term investigation into the effects of explicitly sexual chats and the emotional bonds that this type of interaction can create in users. They point out that at the moment there is no “empirical evidence” about the impact, but for now they are clear. And yet, there is another great reason to cancel it indefinitely.
Let’s focus on what makes money. In recent weeks we have seen how the new pretty girl of the world of AI is Anthropic, which with Claude has managed to conquer precisely the market sector that is beginning to generate income decent for AI: the companies. OpenAI had been especially focused on end users, but the steps it has taken to try to convince us to pay for ChatGPT Plus/Pro They don’t quite work.
No ads, no shopping. A few months ago OpenAI announced that ChatGPT was already capable of buy things for you with its Instant Checkout. The feature was really promising and proposed a paradigm shift in the rules of traditional e-commerce, but this launch seems to have had much less impact than expected. The decision to place ads during conversations seems not going to make ChatGPT’s revenue skyrocket either, so the solution is becoming clearer: if we have to be like Anthropic, we will be more Anthropic, we imagine Sam Altman is saying.
Goodbye Sora… The ads don’t quite work, neither does Instant Checkout, and many other launches have not gone beyond generating a fleeting expectation. It happened with Sora: that OpenAI I abandoned her It is a disturbing sign that the company prefers to completely recalibrate.
…hello superapp. Another sign of this reorganization is the fact that OpenAI is preparing a desktop tool that will unify its chatbot, its code platform (Codex) and the Atlas browser. The objective, to create a super app with agentic capabilities, not only oriented to code, but also to productivity. It is not clear if they will launch it as a solution for end users or the destination will be the company, where Anthropic is winning the game.
New ‘Spud’ model in sight. In The Information indicated this week that OpenAI had recently completed development of a new AI model called Spud. OpenAI is expected to launch it in the coming weeks, and Altman reportedly told his employees that such a model “can really accelerate the economy.” It is not clear what it refers to (agent capabilities?), but with it OpenAI may be able to regain some of the ground lost with Anthropic. If Anthropic lets him, which we doubt.
Image | Universal Pictures
GIPHY App Key not set. Please check settings