JUHASZ: DiBenedetto now works for Louisiana’s Department of Education and is in charge of bringing Amira into more classrooms. He says by the time the state’s two-year pilot is over…
DIBENEDETTO: I think we’re going to see some interesting impacts, and we’ll definitely have some data to make prudent decisions in the future.
JUHASZ: Like whether to spend even bigger money on AI. The company behind Amira says 2 million children already use the tool. Experts caution the technology isn’t a replacement for teachers or even all tutors. It can’t build relationships with students like humans can.
MONTAGNINO: I’m old-school. I still believe people, especially with reading for little kids - that’s where it’s at.
JUHASZ: Montagnino, the principal in Gretna, says for that reason, she was skeptical at first.
MONTAGNINO: But this, to supplement good science of reading instruction in the classroom? This is great.
JUHASZ: And it’s likely to get better because just as kids are learning from Amira, it’s learning from them, too.
[Bolding added]
So it seems an alternative headline for this story would be “Private for profit company gets paid to collect training data for its AI from children who could face disciplinary or legal consequences for non-compliance”
helping children learn to read sounds like an ideal use case for an LLM. An app that utilizes its own users interactions to enhance its own capabilities is not inherently malicious and is vastly different from selling user data to third parties or training on scraped content from others.
And what are you even talking about with the “children could face disciplinary or legal consequences for noncompliance” nonsense. where was that in the article?
“Children could face disciplinary or legal consequences for noncompliiance” is just how school works.
Exactly, they’re a captive audience, and moreover they are legally incompetent to consent to a contracted business relationship like this
If this was a department of education AI or even some kind of transparently administered non-profit organization I’d be fine with this, but the fact that this is being developed for some for profit company that can just jack their rates and cut off public schools whenever they want to is bullshit. Like, I’m not opposed to the technology of LLMs at all, I think they’re actually pretty neat, but our social and economic systems have a lot of exploitative trash in them that cool technologies can inadvertently exacerbate.
Do you think the department of education writes the textbooks, standardized tests (SAT, ACT, etc.), grading and student management software, learning management systems (Google Classroom, Canvas), or manufactures its own classroom tech (Chromebooks, tablets)? The education system is full of for-profit businesses that can jack up the prices, and they do. The DOE simply doesn’t have the resources to create these things themselves and would cost them far more if they tried. The only new thing here is the AI, the business model has existed forever
Personally, I’m more concerned with the use of Google products in schools. A company that’s sole business is harvesting user data and selling it to advertisers should have no place in schools or children’s products. But they’ve embedded themselves into everything so people just accept it at the cost of privacy
There is a difference between the old fashioned scams with textbooks, and the new AI scams. Its true they have a lot of things in common. But the AI scams are more centralized, more powerful, can make more money. These are forging new ways to con people.
But we have gone through several generations of new scams, they are not all bad. Such greed and kickbacks are probably healthier than realized and contribute to a stable society somehow in a way that escapes me, but I feel is valid
Each one of those has a bunch of particular nuances, but in general - yeah, I think they could and should in a lot of those cases
Yeah, it’s a big problem with a lot of little parts to be tackled
Then government should give them the resources (actually, I think a whole separate agency that develops open source software for any government agency or anyone else who wants to use them should be established, but that’s kind of besides the point).
I don’t think that’s true, and even if it were I think we should be willing to pay premium to make sure essential systems that support the public good are being administered in democratic ways (e.g. by public agencies that are required to give public reports to elected lawmakers and be subject to citizens’ FOIA requests).
A lot of stupid ideas hang on for a really long time. Like, we still have monarchies in the 21st century world.
I 100% agree this is a significant problem too, I just haven’t come across any good articles about it recently