What is OpenAI's dots assistant?
OpenAI has released dots, a personal assistant built to keep working in the background whether or not a user asks it to. The company introduced dots at its developer day in late September. It is available in the ChatGPT desktop and mobile apps to subscribers on the Pro tier or higher, with plans starting at $100 a month.
Alexander Embiricos, who leads OpenAI’s enterprise product team, explained where the product came from at TechCrunch Disrupt this month. He said the assistant grew out of work the company was already doing on ChatGPT Work, which itself came from the agentic coding tools OpenAI had built for software engineers. The team’s goal was to repackage those capabilities into something more accessible. Even the name is deliberate. It is a play on the negative space inside OpenAI’s logo and is meant to “feel like just something that could eventually be ubiquitous.” Asked when he learned that Elon Musk owns the dot.com URL, Embiricos said, “I found out very late.”
Once connected to a user’s calendar, email, computer and other tools, the agent tries to help get work done. Embiricos acknowledged that much of dots’ functionality can be reached through Codex, OpenAI’s coding tool released at the beginning of this year. That route, however, takes more effort and more hand-holding from the user.
How does dots' proactive research work?
The defining feature is that the assistant does not wait to be prompted. Embiricos described it as “sort of thinking about you and thinking how to help you.” OpenAI calls one piece of that proactive behavior “proactive research.” On a regular cadence, the system reviews what it knows about the user, the conversations it has had and the resources it has been given access to. It then looks for ways to be more helpful and proposes them.
The early results are promising but uneven. In TechCrunch’s testing, the assistant flagged calendar conflicts and upcoming deadlines, though not always in time to be useful. Embiricos recalled that early in development, it warned him about a conflict just three minutes before the event. “I only have three minutes!” he said. He expects the timing to improve. He also pointed to a story from an engineer at a drone delivery startup, who said their dot had noticed an outage and analyzed it before the engineer did, then brought back information on how to fix it.
Does dots collect personal data?
Privacy is where the tension sits. Any personal agent needs access to personal information to be useful, and that data could be valuable to the company that runs it. Meta’s Muse has drawn criticism for the profiles it builds of users’ family and friends. Asked whether dots collects similar data, Embiricos pointed to proactive research and said the system uses what it knows about the user to decide what to suggest. A separate startup developing its own AI personal assistant, Hark, is explicit that it does not collect data to fuel an advertising business, a distinction it draws from OpenAI and Meta.
OpenAI says dots carries the same privacy controls as ChatGPT. Users can turn off model training, which keeps their conversations from being used to improve OpenAI’s AI, and they can control which resources the assistant can access on their computer. Those controls are not granular, though, and the app can be insistent. It refuses to accept read-only access to cloud drives. Embiricos said the company needs to “build out super, super, super granular controls.” He also acknowledged that the launch drew complaints about the product’s conservative defaults. The assistant asked for permission often. “I remember knowing that we were going to see that feedback,” he said, “but thinking, ‘This is okay, we can tune that over time.’”
Scale is the other open question. Agents are fun when only a few people use them, but businesses may not be ready to deal with them, and many people do not want to deal with them at all. Embiricos has a human assistant, partly because OpenAI’s corporate partners want to interact with a person. “I personally never intend to have a meeting with someone else’s agent,” he said.
The bigger test may come with money. Embiricos said OpenAI is “deep in lots of conversations with payment processors,” while adding, “I probably can’t say too much here.” An assistant that can spend a user’s money would be a lucrative step for OpenAI, but it would also be a major request of consumers. Giving an AI assistant access to an email inbox already feels like a big ask. Handing it a bank account is a much bigger one, and for now, that account remains off limits.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.