If you’re mad at this… do you like, not use credit cards? Every payment processor “uses ai” to determine fraud. - ThreadSky

steveklabnik.com • 20 days ago

If you’re mad at this… do you like, not use credit cards? Every payment processor “uses ai” to determine fraud.

Ensuring user safety at scale takes massive resources. That’s why we’re excited to partner with ROOST—helping smaller orgs enhance safety, meet regulations, and compete. More safety, more competition, more user choice.

Comments

penryu.bsky.social•20 days ago

Have we really lumped in simple basic heuristics in with "A.I."?

Or are we just lumping all soft, unreliable, hand-wavy systems together as "A.I."?

trist.am•20 days ago

I think we are fast approaching the point where advertising a machine learning-based solution as “AI” will become a strictly negative association.

Unfortunately this happens whenever a VC-funded gold rush takes over a term-of-art (i.e. ‘crypto’ now means bitcoin, not cryptography)

penryu.bsky.social•20 days ago

I get the feeling the term "A.I." is coming to mean _less_ in general.

Or possibly, it is coming 'round to mean as little as it ever did.

bitcoinbro.bsky.social•20 days ago

Not quite right. “cRyPtO” is 36 million “coins” or projects that are all affinity scams to “Bitcoin”.

penryu.bsky.social•20 days ago

Generative AI has more in common with spamassassin than with the much more deterministic heuristics credit card companies use.

hiivelabs.com•20 days ago

We always have! 🙂

penryu.bsky.social•20 days ago

I've clearly used the term "we" as loosely as others use the term "A.I."

hiivelabs.com•20 days ago

"We" being the people who actually work/study in ML and have done since before transformer-based networks became the shiny new thing.

Anyway, here's a cat photo.

penryu.bsky.social•20 days ago

Okay, from that perspective, pop culture has done you wrong.

hiivelabs.com•20 days ago

Yeah. It's an annoyance to be sure. AI ≠ GenAI.

steveklabnik.com•20 days ago

Machine learning is AI, no matter if it’s “basic” or not (and in my experience you are incorrect, though maybe we mean different things than “basic”)

penryu.bsky.social•20 days ago

Taken another way, this is the best case I've heard to insist:

Specify the _type_ of "A.I." you mean, or we'll hand off national defence to this Bayesian dataset I have on my 20yo Linode.

penryu.bsky.social•20 days ago

We may be at an impasse here.

I realize that historically, CS has made this association. However, in light of better vocabulary, I feel like this conflation of terms far more harmful than helpful.

eikopf.dev•20 days ago

incomprehensible societal damage has been done by VCs describing literally any kind of statistical inference as AI

differentsmoke.bsky.social•20 days ago

"incomprehensible societal damage has been done by VCs"

Art is knowing when to stop, my friend.

steveklabnik.com•20 days ago

Oh yeah I blame the VCs for this for sure

bumblefudge.com•20 days ago

it's not "artificial intelligence" unless it reduces payroll, otherwise it's just sparkling brute-force statistic inference. it's not VC's fault, it's the [anti-human] nature of unfettered stakeholder capitalism

bumblefudge.com•20 days ago

until corporate taxes go back up, there's no real incentive to hire humans, and activist shareholders will step in any time someone tries using "natural intelligence" instead of the cheaper, enshittifying artificial variety. all of this is downstream of tax policy and capital mobility

hiivelabs.com•20 days ago

It's not that people are stupid; it's that they've been conditioned (one way or the other) to have a knee jerk reaction to the term AI, whether it's GenAI or not.

hiivelabs.com•20 days ago

(And they're stupid for letting that knee jerk reaction get in the way of facts, but that's another issue — the "All AI Bad" people are just as ridiculous as the "AI everywhere!!!" Techbros. Neither group understand what they're talking about.)

differentsmoke.bsky.social•20 days ago

Respectfully, this seems like a willful misunderstanding of the context in which this issue arises. Yes, this may not be AI in the generative sense. But also, it is launched at a conference that (and with branding designed to) takes advantage of the ambiguity of the term AI.

differentsmoke.bsky.social•20 days ago

Best case scenario, even if this is a 10/10 Machine Learning tool that really helps people, the choice of branding helps launder the reputation of the intentionally ambiguous "AI field" as a whole.

mariaoconnor.bsky.social•6 days ago

What about creating in each blue state, a state owned bank, with lower late pmt penalty & credit card interest rate? At least 1% of the gains should go to state medical system, state education system, low cost housing & some type of state retirement system for the people of each state. 4% total.

floracatz.bsky.social•20 days ago

It doesn’t help that ROOST’s site and his comment are kinda vague on what problems it’s solving. I spent way too long last night trying to understand what concrete solution was being purposed, and just threw my hands up and guessed it was automated moderation.

floracatz.bsky.social•20 days ago

It’s wild that people think moderation should be done purely by people. Why in the world would you want someone to be forced to see humanity at its worst 40 hours a week and expect that not to harm you. Facebook human moderators reported having trauma.

aaronblackshear.bsky.social•20 days ago

No they don't. The term has been stretched beyond meaning

mstrofbass.bsky.social•20 days ago

When I was a CS undergrad in the early 2000s, we were calling basic neural net research “AI”. The stuff we have now is magnitudes better than what we had then.

How has it “been stretched beyond any meaning” if it’s more applicable to what we have now than then?

aaronblackshear.bsky.social•20 days ago

Basic statistical inference is not "AI". I don't care what someone in the past did. Most of the things being called "AI" today are not, the term has lost all meaning. People are using it because it is the new buzzword, not because it is anything even approaching artificial intelligence

mstrofbass.bsky.social•20 days ago

The point, which you so eloquently failed to get, is that it is you who is redefining the term. Everyone else is using it properly.

aaronblackshear.bsky.social•20 days ago

Exactly backwards. People have used it wrong for years, in less egregious ways, and now they are using it for basic statistics, rendering the term meaningless

mstrofbass.bsky.social•20 days ago

No, you don’t get to come in decades later, redefine a technical term of art, then get mad that everyone is using it as they always have.

cassidymoen.bsky.social•20 days ago

Now that you mention it, things probably were better overall when it was easier to do a little credit card fraud to get through the month or for an enterprising individual to improve their station in life.

without.boats•20 days ago

Theyre just stupid, Steve

steveklabnik.com•20 days ago

I originally thought so but I am seeing people who I don’t believe to be stupid posting like this

without.boats•20 days ago

Then they’ve been polarized into taking stupid positions instead of considering reality

luteart.bsky.social•20 days ago

I don't think it's entirely the fault of the individual though. Tech firms immediately bandwagoned on calling *everything* AI, from nuanced machine learning to actual harmful shit like GenAI, because it's the shiny new marketing term that's a honeypot for investors.

The word has lost its meaning.

without.boats•20 days ago

generative AI is not harmful lol

steveklabnik.com•20 days ago

back in 1965 people were saying the same stuff sam altman says now (they were wrong then and he's wrong now)

"has the definition of AI lost all meaning" has its own wikipedia page https://en.wikipedia.org/wiki/AI_effect

conroy.org•20 days ago

these people have obviously never had to build trust and safety tools

jeremynt.bsky.social•20 days ago

I think it's a terminology issue more than anything. People rightfully don't want somebody asking ChatGPT if a transaction looks legit and don't think of other ML the same way now.

crystaldawnmf.bsky.social•20 days ago

I’ve seen companies use “AI” to describe what is essentially spellcheck. The term means nothing to the general public because too many corporations wanted to bandwagon onto the new shiny thing even if they don’t actually provide said thing.

jsstevenson.bsky.social•20 days ago

Wait til people learn about the weather report

destructuring-let.bsky.social•20 days ago

imo people just love screaming at aaron “thanks for the heads up” rodericks, the head of trust & safety at bluesky dot social

slime.bsky.social•20 days ago

they want humans to do it; artisan crafted payment analysis by hand with skilled fraud detection. manually done with love.

rosactrl.bsky.social•20 days ago

And looking for another angled: I wish no human had to do that work. Ever

simon.farnz.org.uk•20 days ago

Unfortunately, I know a few "people" who consider the idea that AI might reduce the number of humans in content moderation roles a problem, because they believe that making the low-waged suffer through abusive content is a virtue.

steveklabnik.com•20 days ago

ugh

rosactrl.bsky.social•20 days ago

From another angle* geez

steveklabnik.com•20 days ago

yeah this part is rough too

gienieczko.com•20 days ago

You are right BUT ALSO this is entirely tech industry's fault for so utterly poisoning the term AI that you just instinctively cringe whenever you see it.

If OpenAI introduced an "AI-powered safety tool" they'd use the same words, but meaning they feed your posts to ChatGPT and ask if it's fine.

gienieczko.com•20 days ago

Now every time you want to talk about something that, you know, actually works, you have to say Machine Learning or the public will assume you mean the wisdom woodchipper

mcoorlim.com•20 days ago

It's probably the involvement of OpenAI that has people assuming the worst about ROOST

steveklabnik.com•20 days ago

Or like, do you turn off your email provider’s spam filters?

cursed.monster•20 days ago

Bayesian probability is inherently sinful

cursed.monster•20 days ago

(yes, I know, it’s not actually how we do spam filtering anymore)

steveklabnik.com•20 days ago

Lmao, Anti-LessWrong-Aktion

steveklabnik.com•20 days ago

Like, if you check what the weather forecast, well, guess what?

A lot of that is physics based, but it is very often blended with AI

steveklabnik.com•20 days ago

I fully agree that the industry’s turn to “put AI in everything” is pretty cringe and LLMs are super oversold but like… this stuff is *old*. Including content moderation!

immortalscience.org•20 days ago

we would have just called it “machine learning” before companies needed to hit the ai square on investor bingo cards

steveklabnik.com•20 days ago

hilariously in the 90s, it was the opposite of today: AI folks had to take "AI" off of what they were doing, because it was seen as a negative not a positive

Comments

Posting Rules

Reply