Using sequences of life-events to predict human lives

by SubiculumCode • 3 years ago

98 51 3 years ago

I wonder if fine tuning GPT-4 on the records would do significantly better

anigbrowl3• 3 years ago

I have mixed feelings about this. It could be a great tool to help people, but on past experience it seems more likely that HR departments/ financial service providers/ governments would use it to write people off if they have too many negative life indicators.

reactordev• 3 years ago

This. It would be the envy of every insurance company. How risky are you really? I’m glad I’m not the only one that immediately thought about this.

matrix87• 3 years ago

Insurance companies are probably already doing something like this anyway

peyton2• 3 years ago

They’re doing it anyway. Sunlight is the best disinfectant.

MattDaEskimo• 3 years ago

I'd like to believe that lives are way too nuanced to properly train and evaluate a transformer this way.

Also, I would wager that upon release, suddenly everybody says they eat healthy and exercise daily.

shadowgovt• 3 years ago

I have an acquaintance who is convinced that Google ads stalk him because the ads make inferences yet shouldn't be able to, like offering flight prices from one city to another he frequents when he's never done that search in Google before.

I've tried to explain to him that people aren't actually very unique in large models, and there's probably several signals Google does have from his other searches to conclude he's interested in those two cities.

wharvle• 3 years ago

Every time we get better at things like this, all the power tends to stay with organizations, and individual people are lucky if they get anything out of it.

Amazon, Wal-Mart, MacDonalds, giant banks, insurers, et c., analyze their competition, suppliers, and consumers in great detail with computer data and models. Where’s the widespread automated buyer’s agent for consumers, to balance that power so strongly that it shoves profit margins the other direction, so it’s not just megacorps pushing ever closer to some absolute limit of the last tenth of a cent they can wring out of us? Fucking nowhere, they’d never allow that kind of access to the data we’d need, the best we get is “CamelCamelCamel”, not some Siri-alike we can casually tell to go find the best price/quality intersection on the market for [product] and have that show up a couple days later.

And we’ll never get it. Just an ever-tightening noose of perfect price discrimination and risk calculation that goes only one way. Short of legislating radical levels of availability (and reliability and consistency—standards and minimums) for price and product information, anyway, and I’m pretty sure that’s on nobody’s radar.

zmgsabst• 3 years ago

Scraping is legal; as are assistive technologies.

What’s lacking is the collective willpower to develop social organizations, which could orchestrate and fund such efforts. There’s a complete lack of organization among the public.

The problem is neither technical nor legislative: people just don’t care to, and are happier accepting the status quo.

otikik• 3 years ago

Seize the means of computation

Aerbil313• 3 years ago

And here I am just watching Kaczynski’s predictions all the way back at 1970 become truer with the decade. Technology clashes with human autonomy.

It’s even argued that the path of technological progress would not change even if you there were aliens, AI bots, any entity with the instinct of self-preservation, intelligence and replication. The mechanisms of the macro-system comprised of human biology + technology are explained in his book Anti-Tech Revolution in great detail.

It was a hard and bitter conclusion to come as a software dev, but I can never again unsee the fact that all technology is inherently evil. For a shorter but enlightening read, see his original manifesto “Industrial Society And Its Future”: https://www.washingtonpost.com/wp-srv/national/longterm/unab...

derbOac• 3 years ago

This is a critical problem of our time in my opinion. Persons are approached as static unchangeable objects, as if their trajectories are deterministic and what you're doing with these models are resolving the trajectories like a telescope or microscope. Measuring the substance of the person so you can extract from them.

The other schema is one where you're modeling manipulation effectiveness, to probe and predict how you can best (most effectively with the least resources) achieve some change in the trajectory. Engineering a person. It seems like we really only think of interventions as trying to return someone to some baseline condition, rather than to try to improve people beyond the baseline, to invest in them.

The paper has a bit at the end about using it for personalized interventions but it feels performative, like lip service. I know one of the authors professionally in a distant sense so I think they're probably genuine about it, but I can't help but wonder how the paper might be different if intervention was actually their focus. Wouldn't they be modeling changes in trajectories, the deviations or something?

reality_inspctr• 3 years ago

I tried to build a startup based on a similar idea, but the traction was not good. The squarespace website is still up https://sundialcalendar.com/

It’s stunning how much of our lives are predictable. We are free will surfers on waves of determinism.

walterbell2• 3 years ago

We are free will surfers on waves of determinism

As the cost of real-time surveillance, prediction and "nudging" continue to fall, the cumulative feedback loop can converge into human archetypes which are as much a function of the observer as the observed.

In an ocean of sensors and stimuli, how can past free will influence our future selves? Any stories you can share from users of your service?

pinkmuffinere• 3 years ago

The website is still up, but I take it the startup is fully shuttered? If not, I'd love to try it. Clicking on the Signup/login button I see a message saying:

This Deployment has been disabled.

Your connection is working correctly.

Vercel is working correctly.

benob• 3 years ago

I wonder how an ethical board was convinced to allow this kind of research

karencarits• 3 years ago

It's surprising to see that this was done in Denmark, on data on more than the million people, and I couldn't find any reference to the ethical board/approvals in the manuscript

herval• 3 years ago

what's unethical about it?

fho• 3 years ago

These days every time you do experiments that involve humans it's expected that you run your experiment design by your institutions ethics committee.

In this case I would expect that the committee perked up their ears, but once they heard that the data was anonymous they did not have any further concerns.

samstave• 3 years ago

When I was a kid, I was involved in a study with UCSD. I grew up in California on a Commune in the 1970s very famous in the Bay area. The study with UCSD would follow a range of various people from all various socio-economic backgrounds.

I first recall going to San Diego when I was 3. They would check in on me every few years up until I was 18 and I asked to leave the study.

I don't recall what it was named... but it was interesting. I wont share the insights they had of me - but some of them were insulting and very wrong.

This will do that at scale.

m3kw9• 3 years ago

Didn’t see input for astrology sign though, maybe they can make it more accurate lol

greatpostman2• 3 years ago

One of my rules for life:

If someone does something once, they will do it twice. Given similar circumstances, people act the same

amelius• 3 years ago

Does that hold for mistakes too?

greatpostman2• 3 years ago

Ofcourse

resource0x• 3 years ago

Unfortunately, the rule still holds when "someone" is yourself. :-)

IAmGraydon• 3 years ago

Well...yes. We aren't random action generators.

walterbell2• 3 years ago

Predictive algos have long been used for market segmentation, predating LLMs and were illustrated in WestWorld S3E3 screenshots of individual human lifespan profiles, https://gizmodo.com/the-tech-in-westworlds-episode-3-shocker...

SCUSKU• 3 years ago

Part of me wants to be able to encode my life2vec and see what the next predicted tokens are. On the other hand, as much as going to a fortune teller sounds fun to me, I worry that if I go I will simply make what they say come true since I will be anxious about what the fortune teller said.

That said, my life is already on a certain track, and because I'm afraid of change, it's unlikely to be much different than it was before. Maybe this is just a weird call to actually do what your gut says, and to follow your dreams.

dkjaudyeqooe• 3 years ago

because I'm afraid of change, it's unlikely to be much different than it was before

Life changing events are often out of your control

jedberg3• 3 years ago

That's sometimes but not always, or even often, true. You're thinking of accidents or death. But what about marriage or jobs or education or divorce? Those usually require an active choice (or many of them).

p-e-w2• 3 years ago

Our work has important societal and ethical implications, which we outline in the Discussion as well as in Methods, Sec. 4.1, and SI: Model Card.

Maybe those ethical implications should have led to the decision not to pursue this avenue of research at all, or not to publish the results.

These "ethics disclaimers" are so stupid. It's as if scientists believe that they can somehow free themselves from moral responsibility by waxing philosophical about the ethical implications of their research in an appendix.

That's not how it works, folks. By choosing to publish, you have already incurred shared responsibility for whatever is built upon your work. Claiming that you have "considered" this means Jack Shit.

peyton2• 3 years ago

Strange to see these ideas on a site called “Hacker News” given the “hacker ethic” is to make information available to all: https://en.m.wikipedia.org/wiki/Hacker_ethic

p-e-w2• 3 years ago

Like other vague, vogue philosophical concepts, the Hacker Ethic is many things to many people.

To me, "every avenue of research must be pursued at any human cost, and ethics are at best an afterthought" has nothing to do with being a Hacker.

Animats• 3 years ago

So that's how Precrime works.

Jimmc414• 3 years ago

I believe this is the corresponding code:

https://github.com/SocialComplexityLab/life2vec

sinuhe69• 3 years ago

“the most comprehensive registry data in existence, available for an entire nation of more than six million individuals across decades. Our data include information about life-events related to health, education, occupation, income, address, and working hours, recorded with day-to-day resolution.”

I shuddered just reading this. What a nightmare for individuals’ freedom and privacy. What a bloody nation is that?

woleium4• 3 years ago

How do we build a system where there is enough trust that this data will be used for “good”?

manofmanysmiles• 3 years ago

Address the problem that people have a proclivity to do "bad". Call it mental health, resolving trauma, spiritual practice. When people are no longer wounded and "broken" internally, they will go out and sculpt the outer shared physical world to match their improved internal state, and all will be good.

I'm not sure this is possible, but I am sure that any attempt to control human behavior externally will fail.

Using sequences of life-events to predict human lives

Related Stories