Dr. Claire Brooks

President @ModelPeopleInc. #GlobalConsumerResearcher. #Author of #Marketing with #StrategicEmpathy #MRX #Insights #strategy Founder member @worlddignity

[Dr. Claire Brooks @modelperson]

There is no societal permission for an AI future that hollows out entire industries.

[Elon Musk @elonmusk]

Interesting

[Elon Musk @elonmusk]

Interesting

[Dr. Claire Brooks @modelperson]

@thetrainline my train was 27 minutes late outbound, 40 mins late return. Your email catfished me to open an account for a sizeable refund...of 60p. Misleading comms using my personal email is unprofessional

[Dr. Claire Brooks @modelperson]

Wonderful ethnographic analysis on the #keirstarmer entity

[Tim Shipman @ShippersUnbound]

This is the best piece on Starmer you will read today, perhaps this year

[Dr. Claire Brooks @modelperson]

@Avis_Worldwide_ Done. What's the answer please?

[Dr. Claire Brooks @modelperson]

@avis how do I resolve a fraudulent car rental bill from your San Luis Obispo location? I have spoken to 6 phone reps who were all completely unable to help

[Dr. Claire Brooks @modelperson]

Celebrating 10 years of Strategic Empathy https://t.co/W5E9A8KdQj

[Dr. Claire Brooks @modelperson]

This reflects my recent experience at @DFW. Last minute gate change, on app but not on board. A 10 gate walk, helping an elderly lady with a cane. We only made it bc flight was delayed. Better coordination needed @AmericanAir

[ac @auraxchains]

@AmericanAir also on my flight here you guys didn’t update your boards or the website to where our gate was so we had an entire flight of people standing around confused and then it finally updated as we were boarding

[Dr. Claire Brooks @modelperson]

Limitations of synthetic #survey #data. Thanks @verasight_data #marketresearch

[Verasight @verasight_data]

Synthetic data performs best when attitudes are predictably polarized by demographic groups.

In Verasight/@gelliottmorris’s comparison of a nationally-representative survey of 2,000 Americans with a synthetic sample, political questions showed the lowest error, with a mean absolute error of 12.1 points, while health care questions reached 23.4 points. Topics with strong demographic polarization were easier for the LLM to approximate. Questions about lived experience and personal behavior were not.

If your research depends on precision or measuring nuance, how much error is acceptable before the data becomes unusable?

https://t.co/5M2OmlmfD6

[Dr. Claire Brooks @modelperson]

@AmericanAir @EsqHardy Or you post rolling delays to manage crowd fury at the desk?

[Verasight @verasight_data]

One finding stood out immediately in Verasight’s/ @gelliottmorris latest synthetic sampling white paper: LLMs fail at multi answer survey questions.

Across 77 multi response options in our study, 35 percent were selected by more than 5 percent of human respondents but by fewer than 1 percent of synthetic responses. Even more striking, 31 percent of valid options were never selected at all by the LLM, including several chosen by over 30 percent of real respondents.

https://t.co/5M2OmlmfD6

[Dr. Claire Brooks @modelperson]

@Avis I have now had 2 calls from people pretending to be agents asking for credit card numbers and ID. Your organization is being hacked.

[Dr. Claire Brooks @modelperson]

@Avis 01877408US1

[Dr. Claire Brooks @modelperson]

@Avis I have had second in a row scam transaction from your #DFW location. I refueled car. Agent did not ask if I refueled and she refused a paper receipt. They added $16.99 to email receipt. This only happens at DFW and I use Avis worldwide Investigate & refund again please!

[Dr. Claire Brooks @modelperson]

@avis apparently your refund system is down🤨Your stated rep called me on WhatsApp and tried to get me to upload government documentation so I ended the chat. Please refund the fraudulent gas charge to my credit card immediately.

[Dr. Claire Brooks @modelperson]

@EnquiryCustomDM No reply from WhatsApp

[Dr. Claire Brooks @modelperson]

@EnquiryCustomDM How do I DM you

[Dr. Claire Brooks @modelperson]

Well articulated @BrianRoemmele. And true creative content will ne maimed and stolen in the process #AI

[Brian Roemmele @BrianRoemmele]

“Is it AI slop—or is it real slop”

In the decades ahead few will even care.

Our innate dance with novelty and wonderment will break a leg and never quite heal. We will limp in to the decades ahead fatigued by the endlessness of absurdity fuels by algorithmic pumps of dopamine.

[Dr. Claire Brooks @modelperson]

Training #AI

[Brian Roemmele @BrianRoemmele]

By training AI on the lowest form of human data the sociopathy of Reddit and the Politburo of Wikipedia AI models will 100% mirror sociopathy no matter how many “alignments for safety”.

Ethics and morals come from the training data. It is not negotiable.

I know the solution and I guess it won’t matter until this really happens… and it will.