Perhaps a huge tell about national strategy is the fact that the owner has $10s of millions to loan to the company? US economic structure in post WWII era has increasingly focused on return on capital (and value extraction). How can that compete in long term with an economy which prioritizes reinvestment *in industry*?
Voice of American *capitalists* used to the idea of investing capital to generate rent. Maybe AI can rebalance the market of capital and individual labor.
I found this article frustrating. In my experience, the students who participate in the “co-creation of a contract” would likely also have done their best to comply with Professor induced rules. The challenge is the 10-20% who, stressed and overworked, will be just as likely to ignore this contract as any other. I would have liked to understand how the situation is changed compared to no rules at all…
Comparing with China, I’d be curious about whether their companies face the same return-on-capital forces that US (and to a lesser extent western) companies do. So when we outsource to them, it’s to a vendor who is not needing to promised 10%+ returns. As I understand it, this is because the central government exerts pressure on banks to make loans at low rates, so that capital can be raised more easily for targeted industries.
I agree that onshoring is likely impossible without some sort of contextual reimagining of US industry.
Training data of task completion. See, e.g., robots doing backflips. Presumably there’s an optimal robot for gymnastics but if you start with humanoid form you can train based on many videos of human movement. The alternative - world model sim with physics and loss functions- perhaps ends up being too unconstrained when you add in optimization of the robot form…
Lots of complaints about corruption. It’s worth imagining how you would run a very low income developing country. Remittances and tariffs are the easiest items to tax for revenue. I’d love suggestions for better alternatives…
Curious about the design choice. Why not use the TI parts with integrated microcontrollers rather than two separate chips? Or even a FPGA with integrated ARM9 like the Zynq family?
The obvious question I would have asked: given the concern that this may not be ethical if the brains are still “alive” AND the concern that a brain separated from the body probably doesn’t function these same, why wouldn’t we test things in living monkeys (instead of mice)???
It seems that the likelihood is high that the right animal model would yield superior data???
Further counter argument- It seems that (elite) undergraduate students care about their professors research (versus just teaching). Else Harvey Mudd would be much much harder to get into compared with MIT?
To the people who are thinking about other languages, "Lexical decoding" - recognition of a word during reading was the strongest predictor of reading compression (as opposed to phonological).
Restating the highlighted result: Gc ("Comprehension-Knowledge") had the strongest effect on both lexical and phonological decoding. Knowing a word makes it easiest to comprehend when reading. This is probably completely obvious, but the broader point is that rich conversations with students that involve teaching them lots of words will improve their reading.
Only partially supported interpretation/application - All this business about phonics will only take you so far if the adults in a kids life (including their teachers) are not talking to them richly about a lot of stuff. Asking teachers to do a lot of rote repetition risks cutting out the really important part of school where students are actually building vocabulary. Teachers that use/teach large vocabularies may be unexpectedly more effective at teaching reading.
I’d love to take that bet. My deed (in Texas) states that my lot is subject to the rules of the subdivision which include a number of zoning style restrictions. (They’re called “deed restrictions” and are very common AFAIK.)
The subdivision rules are changeable only with a supermajority vote. I believe the city (Houston in my case) is prohibited by the state from unilaterally changing them.
(I wouldn’t mind more free property rights!!! I find TX “liberty” is often biased towards $$$)
The title really should specify “to teach remotely”. And I think more broadly the context is the dream that widespread internet would make it easier to educate people “at scale” (meaning for less money per student).
So maybe the real question is why we ever expected “teaching at scale” to be effective.
I think that it’s quite clear that for an individual, curious student, the ability to use modern LLMs probably makes the ability to be 1-1 tutored (by a human!) cheaper/better. But I don’t think anyone claims that watching random videos on the internet will be as effective for LeBron James as having a personal trainer focused on him.
It seems like the overriding issue is to understand whether students need to take courses they’re not interested in. If the answer is yes, perhaps we need find ways of having these topics be taught by tutorial…
I think that there are stories but unreported for good reason. I heard from a pastor friend that he showed up to a house where one of his congregants was inside (with ICE outside), and basically asked them for mercy, and the supervisor was like “Actually, there is lightning in the area so we’ll need to go away but we will come back later.”
So not complete rebellion, but a little bit of humanity. Viktor Frankl talks about how there were some “good” guards at the concentration camp.