4 ms·

There's all sorts of stuff wrong with this, but why is an agent giving them a "hot or not" rating? Surely that can't be the only thing necessary to flag them b

There's all sorts of stuff wrong with this, but why is an agent giving them a "hot or not" rating? Surely that can't be the only thing necessary to flag them based on their social media posts. As using here as an example, I down vote things I meant to upvote or vice versa constantly. Imagine being denied a visa based on that.

ShamelessC 3y ago
Not sure if you're aware US immigration is sort of a shit show.
- boomboomsubban 3y ago
  I'm no fan of US immigration but even they must fear "we let in this villain as we fat fingered thumbs up. Our bad."
  - estebank 3y ago
    Which makes it more likely that whatever prevalence of false positives are ignored as long as false negatives are kept low. Meaning that people without recourse are affected disproportionally by something they have no control over.
empath-nirvana 3y ago
> Imagine being denied a visa based on that. The position of the US is that everyone coming to the US is intending to move here permanently, and any sort of visitation visa will be _denied by default_ from many countries unless you can prove otherwise, and the hoops they make you go through can be ridiculous. I talked to someone from Colombia who waited years for a visa interview, spent weeks preparing for it and got denied 5 minutes into the call because they didn't have enough money in the bank at home. No appeal. All they wanted to do is visit their brother for a week. They hadn't seen him for years.
- boomboomsubban 3y ago
  Yeah, that's terrible. Now imagine they checked the bank, found they had "enough" money, then denied them entry because they accidentally clicked the wrong button. The policy is often bullshit, but following it and still being denied would be infuriating.
  - potatolicious 3y ago
    And that's one of the major concerns with tools like these. Besides the conceptual problems of spying on people's musings, when's the last time a heavily marketed "AI" tool turned out to be at all accurate? It's one thing for the government to spy on your social media postings and decide you're a risk to the country, it's a whole 'nother thing for that to be an error because you posted about that "killer" concert or how a movie is "the bomb". Given what we know of government contractors and their inflated claims about AI my bet is that this tool - however justifiable or not in its use - doesn't even actually work as advertised.

ShamelessC 3y ago

Not sure if you're aware US immigration is sort of a shit show.

boomboomsubban 3y ago
I'm no fan of US immigration but even they must fear "we let in this villain as we fat fingered thumbs up. Our bad."
- estebank 3y ago
  Which makes it more likely that whatever prevalence of false positives are ignored as long as false negatives are kept low. Meaning that people without recourse are affected disproportionally by something they have no control over.

empath-nirvana 3y ago

> Imagine being denied a visa based on that. The position of the US is that everyone coming to the US is intending to move here permanently, and any sort of visitation visa will be _denied by default_ from many countries unless you can prove otherwise, and the hoops they make you go through can be ridiculous. I talked to someone from Colombia who waited years for a visa interview, spent weeks preparing for it and got denied 5 minutes into the call because they didn't have enough money in the bank at home. No appeal. All they wanted to do is visit their brother for a week. They hadn't seen him for years.

boomboomsubban 3y ago
Yeah, that's terrible. Now imagine they checked the bank, found they had "enough" money, then denied them entry because they accidentally clicked the wrong button. The policy is often bullshit, but following it and still being denied would be infuriating.
- potatolicious 3y ago
  And that's one of the major concerns with tools like these. Besides the conceptual problems of spying on people's musings, when's the last time a heavily marketed "AI" tool turned out to be at all accurate? It's one thing for the government to spy on your social media postings and decide you're a risk to the country, it's a whole 'nother thing for that to be an error because you posted about that "killer" concert or how a movie is "the bomb". Given what we know of government contractors and their inflated claims about AI my bet is that this tool - however justifiable or not in its use - doesn't even actually work as advertised.