grubb

Apparently they got Meta's blessing https://x.com/charlieholtz/status/2120975128158945374

I know that standard BYU has the lowest tuition in Utah, lower than any state school, provided you are a tithe-paying member of the Mormon Church [1]. Otherwise they have a higher tuition, similar to the in-state/out-of-state changes for most state schools. Does BYU pathway tuition work the same way?

I don't have a ton of experience with Western Governor's, but the fact that unlike many online universities they are nonprofit is a good sign. Of course, whether they have access to quality instructors is an unknown to me.

[1]: https://finserve.byu.edu/students-parents/tuition-fees-deadl...

I doubt it. It seems to have been posted to HN at the same time as the blog post itself was posted. The timing of the blog post seems to have been based on when their paper went up on eprint [0]. As you can see they seem to have submitted the paper on Dec 6 (date on pdf) and it got posted to the eprint server on Dec 9 (which is the typical time frame). At the very least, zero reason for the MSR people involved in the paper to base their release date around Telegram.

[0]: https://eprint.iacr.org/2019/1416

Alternatively using z3 [0]:

  (declare-const a1 Bool)
  (declare-const a2 Bool)
  (declare-const a3 Bool)
  (declare-const a4 Bool)
  (declare-const a5 Bool)
  (declare-const a6 Bool)

  (assert (= a1 (and a2 (and a3 (and a4 (and a5 a6))))))
  (assert (= a2 (not (or a3 (or a4 (or a5 a6))))))
  (assert (= a3 (and a1 a2)))
  (assert (= a4 (or a1 (or a2 a3))))
  (assert (= a5 (not (or a1 (or a2 (or a3 a4))))))
  (assert (= a6 (not (or a1 (or a2 (or a3 (or a4 a5)))))))
  (assert (or a1 (or a2 (or a3 (or a4 (or a5 a6))))))

  (check-sat)
  (get-model)
  (exit)

The model is satisfiable where only a5 is true, anyone can run it themselves and play with it [1].

Additionally forcing a5 to be false by adding the following to the list of assertions:

  (assert (not a5))

And the model becomes unsatisfiable.

Edit: Finally, to make sure there's not some possible solution where a5 is true as well as another assertion you could alternatively add this assertion:

  (assert (and a5 (or a1 (or a2 (or a3 (or a4 (or a6)))))))

And the model also becomes unsatisfiable. So 5 only seems to be the correct answer.

[0] https://github.com/Z3Prover/z3

[1] https://rise4fun.com/Z3/1DMW

Just my two cents: I grew up in Utah and did my undergraduate degree there. The Salt Lake Tribune is a great newspaper (I subscribed while living in the state) and it has been rough watching them downsize and lose some great staff over recent years. I am very much of the opinion that good local journalism is a necessity for holding people accountable. For example: the work the Salt Lake Tribune received a Pulitzer for [0]. Anything to help good local journalism continue and survive is nothing but a win in my view.

[0] https://www.sltrib.com/news/2017/05/07/salt-lake-tribune-win...

I grew up in Utah and was able to visit Moab many times over the years. It is a great little town and the sites and things you can reach from it are awesome. One of my favorite hikes at Arches NP is the Fiery Furnace, which is permit-controlled. I wonder if requiring permits for more of the sites would help address these issues?

Of course, I'm also a big supporter of efforts to increase the size of national parks and monuments in the state as well (despite this administration's recent efforts). Preserving and regulating more wilderness would go a long way to hopefully mitigating this overcrowding issue.

Differential Privacy is definitely an intriguing security concept. However, another recent interesting negative is that there may be a smaller performance gap between Differential Privacy and the stronger security notion of Obliviousness (https://en.wikipedia.org/wiki/Oblivious_data_structure) than previously thought.

Specifically, it was recently shown that the fastest implementations of ORAM (https://en.wikipedia.org/wiki/Oblivious_RAM) are actually in line with the lower bounds of theoretical Differentially Private RAM performance - meaning you gain zero performance advantage by using a weaker security model. Of course, it's unclear if this is true for other data structures and usages, but ORAM is a non-insignificant research area.

See https://eprint.iacr.org/2018/1051 for more details.