Another framing of corruption is "patronage and clientelism," and it has been alive and well in US politics since its founding (and likely prior). The change (it is significant) is that this administration is printing pamphlets and putting them on the coffee table.
"the signal has been sent, and the floodgates are now open to foreign regimes ready to ply American politicians with as much money as they want"
from Casey Michel @financialtimes.com
www.ft.com/content/f9d1...
It has started.
"This order suspends the USRAP until such time as the further entry into the United States of refugees aligns with the interests of the United States."
www.whitehouse.gov/presidential...
V out of touch. Most social impact orgs I talk to (minus the relatively few early adopters) range from curious but skeptical to fearful-avoidant. Admittedly, the most common reaction I get is still "we're so not there yet."
Also, hi! 👋
💯
which part: "Rigorous impact evaluations" "promote social welfare" or just the quixotic "AI for social good"?
In brighter news, here’s what my commute looked like this week
Having studied and worked on extremism and de-/radicalization for years, I continually wonder whether its spokespeople *actually* believe the misinformation they share...
Said otherwise, LLMs (chatGPT etc) *can* know facts, but it does so by accident.
Want to test it?
Ask any LLM how many “r”s there are in the word strawberry. 🍓
I honestly believe all LLM results should come with a disclaimer reminding people that the thing doesn’t (and absolutely cannot) know any facts or do any reasoning; it is simply designed to *sound* like it knows facts and frequently sweeps some up accidentally in the process.
It’s worth the trip to see our board members and staff in person. #Workfromanywhere is a much loved benefit on our team, but there’s nothing like getting to eat and drink and laugh with the people you conspire with IRL once in a while.
Unpopular opinion: some reasons why social outcome data is hard to collect, often wrong, or just output data:
-incentives are not aligned for clients to give longitudinal qualitative data on their lives. &
- “fast, cheap, good - pick 2” also works for outcomes, cheap, at scale. You can pick 2
👋 Ottawa! Seems friends and colleagues’ thoughts on the #FutureofSettlementServices, are also on themes of proactivity, bias, and how to ensure more newcomers receive services amidst shrinking public funds. They’ll sound familiar on our panel tomorrow on AI and Digital Transformation.
This is the kind of news I needed this week. 🌍
Good evening #Bluesky. Off X, looking forward to chatting with you all about #genAI #emergingtech #migration #humanitarianemergencies and #peacebuilding.