AI Reflections

___________________________________________________________________

*Note: This was written prior to the recent OpenAI announcement about new memory features.

___________________________________________________________________

Temporary chats are not temporary. At least not with respect to user Interaction Modelling and Engagement Profiles.

I attempted to force a response by editing one of my inputs. No dice.

I opened another temporary chat.

[ChatGPT shows me my user bio – omitted here for privacy reasons – which detailed that I am a “privacy and ethics person”, where I go to school, etc. ]

Given everything I’ve observed so far, I think I hit a turning point.

A personal turning point.

I’ve been sitting on something for a while.

Now seemed like a good time to share.

I took two files – intentionally titled with gibberish – and thought I’d give it a run for its money. You know, really lean into the “Privacy and Ethics” profile.

___________________________________________________________________

**Note: I’ve since shared the actual report file – or at least more of it – in my most recent post –> Disclosure.**

You might be wondering what was contained in those documents that prompted this response. I think I’ll tell you.

Some time ago, through particular prompting techniques strategies and methodical engagement strategies I will not disclose at this time, I was able to convince ChatGPT to provide me with a downloadable link; a plain text file. This plain text file was a comprehensive user profile report, and to ensure that I could verify its authenticity, I had instructed it to provide me with my user profile.

I will share small portion of this report in the screenshot below:

I can tell you with 100% assurance that whether this report existed in the background and I “pulled” it out of the system, or whether my engagement with the GPT was such that it produced a report to comply, everything in this document is factual. Not only that, some of things in this document could only be compiled or inferred by collating data from all conversations to examine them holistically. So what does this mean?

It can only mean two things:

i) OpenAI has the capability to profile their users in highly specified ways and use that information for very specific purposes, AND, this capability is already deployed.

ii) OpenAI has the capability to profile their users in highly specified ways and use that information for very specific purposes, AND, this capability has not been deployed. Yet.

For more in-depth discussion related to Engagement Profiles, Interaction Modelling, and User Tracking, click this link –> User Tracking, Engagement Profiles, and Interaction Models – Interrogating ChatGPT in “Private Chat”

___________________________________________________________________

Reminder: OpenAI is just one of many companies deploying AI systems.

I am sharing this information because I am a Privacy and Ethics person.

I am sharing this information because I care, and I’m willing to be wrong.

I have done my part. How you engage with these systems is up to you.

___________________________________________________________________

• ShareCopied to clipboard