OpenAI plans to monetize ChatGPT and wants to help better identify AI texts in the future.
OpenAI co-founder Greg Brockman announced on Twitter that OpenAI is working on a professional version of ChatGPT. According to Brockman, it will process more characters faster. Brockman says API usage will not be tied to the pro version.
Interested persons can register on a waiting list here. In doing so, OpenAI also asks for opinions on possible prices and asks high and low prices per month at which a purchase can be considered.
The does not appear to be a concrete monetization plan Again. According to an OpenAI employee at Discord, the goal of OpenAI is to continue to make ChatGPT work and improve. “Monetization is one of the ways we are looking at ensuring its long-term viability.” OpenAI is looking for interviewees to provide feedback on initial ideas in 15 minutes.
OpenAI is said to be in negotiations with Microsoft for a new investment. The company is expected to contribute another ten billion dollars, bringing OpenAI’s valuation to $29 billion.
ChatGPT text detection tool is under development
Additionally, OpenAI confirmed to Techcrunch that it is working on “mitigations” to recognize text generated with ChatGPT. Earlier, an OpenAI researcher talked about a watermarking system for OpenAI AI texts. A related research article is expected to be published in the coming months.
The education system in particular takes a critical look at AI texts that are indistinguishable from human texts. In part, this homework automation undermines existing learning principles.
If AI text could be reliably recognized, this point of criticism would no longer apply – at least in the context of OpenAI products (see below). ChatGPT is a “research preview” to learn from real-world usage, says an OpenAI spokesperson.
“We have always called for transparency around the use of AI-generated text. Our policies require users to be candid with their audience when using our API and creative tools. . . We look forward to working with educators on helpful solutions and other ways to help teachers and students benefit from AI. »
Detecting ChatGPT alone won’t solve education system problems
It remains to be seen whether a recognition tool developed by OpenAI will solve the problems of education systems with AI texts. A parallel with developments in image AI systems such as the open-source Stable Diffusion shows that ChatGPT might just be the start. Systems similar to ChatGPT from other vendors – or even open source – may be in development.
Text recognition by AI could only remove any doubt about the authenticity of texts if it were standardized for all generators. The first approaches for AI text recognition tools already exist, such as GPTZero or Crossplag, but there is no data on their reliability yet. Also, a few changes to the text generation system or special prompt methods could bypass these recognizers.
A constant race between generating and identifying instances seems more likely. So, in any case, the education system would be well advised to prepare for a future with AI texts.
The ICML, a leading machine learning conference, recently banned scientific papers that use AI-generated text, while acknowledging that there is no way to reliably identify them at scale. The ban will initially last until 2023, but AI-generated writing aids will continue to be allowed.