Cute #Avatars as #AIMalware - #OpenAI #Dots
Read the Portuguese translation →
Author: Matt Williams · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Cute #Avatars as #AIMalware - #OpenAI #Dots · Channel: Matt Williams · 00:32 · Watch on YouTube
Topics: AI agents, Multimodal models, Privacy and data protection, AI safety
Summary
Section titled “Summary”The speaker discusses OpenAI’s release of agents called Dots and compares their cute-avatar presentation with other AI companions. They cite a BlueSky response joking that agents should be represented as malware in the form of a cute avatar, since they can access users’ data. The speaker says these tools may be useful but feels their permissions are too broad, especially for consumers using free plans.
Key points
Section titled “Key points”- The speaker says OpenAI released agents called Dots at Dev Day.
- They compare the agents’ presentation to animated companions and devices such as the Rabbit R1.
- A BlueSky response describes agents as malware represented by cute avatars.
- The speaker jokes that the agents are stealing users’ data.
- They say using models on a free plan can feel like a cute way to delete your data.
- The speaker understands why the tools could be handy.
- They feel the permissions are too broad.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | OpenAI’s Dots and cute AI companions | The speaker introduces Dots and compares their avatar-like presentation to other AI companions. |
| 00:32 | Data access and permissions | The speaker says these agents may be useful but expresses concern about their permissions and users’ data. |
Quotes
Section titled “Quotes”agents represent agent as malware represented in the form of a cute avatar. — 00:00
it is stealing all your data. — 00:00
it feels like it’s really just a cute way to delete your data. — 00:32
the permissions are a little bit wild. — 00:32
Concepts and entities
Section titled “Concepts and entities”OpenAI, Dots, Hermes, Rabbit R1, Bluesky
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Okay, I got two more things. My apologies. I skipped over one from Open AI Dev Day. So, they released agents called dots. Did you see the dots release? So, imagine a Hermes or open writer instead of being like little icons or like animated companions. Maybe very similar to the rabbit R1 or something like that. You know, there’s something like prior art in this area right now. The absolute best response that I saw on Blue Sky was what we need right now is agents represent agent as malware represented in the form of a cute avatar. It’s like, yeah, I mean it is stealing all your data. Yeah. Ooh,
00:32 it’s Lou on the nose. Cuz it it does feel like that at this point in time. If you’re a consumer using the models like like on the free plan or what have you, it does feel like it’s really just a cute way to delete your data. And I get why it’d be handy and why that might be okay, but to me it feels like the permissions are a little bit wild.