Colin Fraser, a data scientist, writes on Medium about what large language models do, their limits, and misleading data stories.
Commenters call his piece on LLMs and chatbots a great, engaging read, and cite his post debunking the Target pregnancy-prediction story.
Target and the model that found the pregnant girl (bad counter argument here: https://medium.com/@colin.fraser/target-didnt-figure-out-a-teen-girl-was-pregnant-before-her-father-did-a6be13b973a5
There are three options:
1. Prevent/Stop it: This ship sailed long ago. Not to be grim about it but pandoras box got opened.
2. Fight it: Tool up, change your print, your behavior, your place. Build focused VM's that you use per topic. Simply do a WHOLE lot less. In the grand scheme, its a lot of work for low return. Note: there are exceptions.
3. Increase Noise: The whole point of most data collection is to sell more to you. Because most people are sheep, a fairly simple model can be surprisingly accurate (over targeting is an issue). Don't be a sheep, diversify, make more noise in the system, search out side your comfort zone and change it up often.
Who are we talking to when we talk to these bots?
https://medium.com/@colin.fraser/who-are-we-talking-to-when-we-talk-to-these-bots-9a7e673f8525
"It is an intentional choice to present this technology via this anthropomorphic chat interface, this person costume. If there were an artificial general intelligence in a cage, it feels natural that we might interact with it through something like a chat interface. It’s easy to flip that around, to think that there’s a technology that you interact with through a chat interface that it must be an artificial general intelligence. But as you can clearly see from interacting with the LLM directly via the Playground interface, the chat is just a layer of smoke and mirrors to strengthen the illusion of a conversational partner."
It’s a meme that will never die, but there’s no proof it ever happened:
> This story doesn’t even show that Target tried to figure out whether the girl was pregnant. It just shows that she received a flyer that contained some maternity items and her weird dad freaked out and wanted to talk to the manager. There’s no way to know whether the flyer arrived as a result of some complex targeting algorithm that correctly deduced that the girl was pregnant because she bought a bunch of lotion, or whether they just happened to be having a sale on diapers that week and sent a flyer about it to all their customers.
https://medium.com/@colin.fraser/target-didnt-figure-out-a-teen-girl-was-pregnant-before-her-father-did-a6be13b973a5
>That's quite an assumption to make. The very famous example of data tracking and creepy correlations is the story about the supermarket sending baby ads to a household based on nothing more than their purchasing habits.
While a fun anecdote, it really shouldn't be taken seriously as something that actually happened.
https://medium.com/@colin.fraser/target-didnt-figure-out-a-teen-girl-was-pregnant-before-her-father-did-a6be13b973a5
The original article : https://www.nytimes.com/2012/02/19/magazine/shopping-habits.html
A counter-claim : https://medium.com/@colin.fraser/target-didnt-figure-out-a-teen-girl-was-pregnant-before-her-father-did-a6be13b973a5
Sorry to have distorted the truth. The AD still revealed to a father that his daughter was pregnant. AFAIC this is relevant to the OP question about why ads should sometime refrain to "target"
Here's a skeptical write-up about the mailers I was referencing. I like this view because it acknowledges human fallacies. For example, that an untargeted ad could feel targeted or that we may exaggerate the amount of ads that feel targeted.
https://medium.com/@colin.fraser/target-didnt-figure-out-a-teen-girl-was-pregnant-before-her-father-did-a6be13b973a5