Skip to main content

I noticed a new feature pop up in Amplitude Data -- “suggestions” on event names, property names, and descriptions.

Is this an implementation of ChatGPT or a GPT variant of some sort?

If so, would you please point me to where in the “fine print” we automatically consented to have our data used with a GPT variant?

Other SaaS companies force you to either consent or reject to this use of your data before making the feature live in your account, which would be good form, given the complicated tangle of ethics and privacy issues at play in the generative large language model side of the industry.

 

Also… My company would like to opt-out of giving our consent and request the means to do so.

Hi @frank.serafine 

That’s a fair point.
Looks like they have started testing their GenAI features for some of the users out there.

From what I saw, the suggested names output ( for saving charts ) seem to be based on existing event / properties/ cohort names used in the chart. So not sure if this explicitly needed to be trained via a GPT-ish approach or via a simple rule based completion.

@jacobn Are you the right person for this from Amplitude’s Product Team ?

CC @Product Team 

Thank you for your question, @frank.serafine

The new "Suggestions" feature that you are asking about uses OpenAI technology to help suggest event names, properties and descriptions that you may want to use, but does not involve sending any of your Customer Data to OpenAI.

Moreover, any AI functionality that Amplitude releases that is powered by OpenAI will use OpenAI's API so that no submitted data will ever be used to train or improve OpenAI's models and all submitted data will be deleted within 30 days of submission.

That being said, if you would prefer that this feature no longer be available in your org, please let us know and we can turn it off for you.

Thanks for the fast response, @jacobn. 

Even in this specific use case where 1) no customer data is sent to OpenAI, 2) no submitted data will ever be used to train or improve OpenAI’s models, and 3) the submitted data will be deleted within 30 days, OpenAI’s business model is very plainly toxic and unsupportable.

They have:

  • Refused to disclose their training data corpus and methods (this is not normal to refuse in data science at large)
  • Refused to state that their training process did not involve copyright infringement and illegal use of others’ data without consent (and have lawsuits against them for these reasons)
  • Developed their algorithm using a contractor based in Kenya that barely paid its workers enough to live ($2/hr equivalent, deep-dive covered by Time magazine)
  • Are under investigation by the FTC over producing misinformation and poor data security

With a variety of competitors in the generative large language model space (hot competition now), both open-source and not, it’s galling that any business chooses to use OpenAI’s services, knowing these conditions. 

Not opting out is tantamount to supporting OpenAI’s business, so please continue turning off the feature for my organization, as we do not wish to support such a business.

@frank.serafine thanks for sharing this, I’ve sent you a DM as well but we’re turning this feature off for your organization. Will send confirmation as soon as it’s done!

What are "Suggestions" in Amplitude Analytics and Data, and how do they function within the platform?

Reply


Terms & ConditionsCookie settings