OpenAI has introduced a new feature on ChatGPT that tracks every click and keystroke made on a MacBook, allowing the AI model to fine-tune its performance. This feature, called ‘Computer History’, enables ChatGPT to learn from a user’s activities and provide more personalized assistance.
The update was announced by OpenAI’s product and engineering manager, Ari Weinstein, who explained that it allows ChatGPT to understand how users work and suggest skills and automations based on their computer usage.
According to Weinstein, the feature lets ChatGPT learn from everything a user does on their computer, including tasks they’re in the middle of and skills they use. A demo video on YouTube shows a development experience engineer, Dominik Kundel, explaining the feature and demonstrating its capabilities.
Kundel asks ChatGPT about the last document he viewed, and the AI model responds with the correct answer. He also asks if the document was shared, and ChatGPT double-checks before responding.
While OpenAI assures users that the feature works on an ‘opt-in’ principle and won’t access content viewed in incognito mode or private browser tabs, some users have expressed concerns about the potential risks to their privacy. They allow users to manually blacklist apps from being tracked and view a data bank of the chatbot’s memory.
However, the fact that the feature can collect sensitive information and isn’t encrypted has raised concerns among some users, who worry that other programs may be able to access this data. OpenAI warns that “Computer History files can contain sensitive information” and may be accessible to other programs running on the user’s macOS.
Some users have compared the feature to Windows Recall, a Microsoft feature that takes screenshots of a user’s desktop and uses AI to auto-scan text. Windows Recall faced controversy when users discovered it could record sensitive information, such as bank statements.
OpenAI has sought to reassure users that the feature is designed to improve the performance of ChatGPT and provide more personalized assistance. However, the company’s lack of transparency in the past has led some to question whether the feature may “accidentally-on-purpose” view details that it’s not supposed to.
As the use of AI models becomes more widespread, concerns about privacy and data security are likely to continue. OpenAI’s introduction of the Computer History feature has sparked a debate about the balance between personalized assistance and privacy, and it remains to be seen how users will respond to this new feature, which may be related to long-term AI missions that require vast amounts of data.
Leave a Reply