andrewparker

OP point out that OpenAI used the "too dangerous to release" marketing ploy with GPT-2... Positioning this as "both sides" have played this card.

But at this time Dario was at OpenAI and was a co-author on the GPT-2 research paper announcing the model.

The "too dangerous to release" approach has been him the whole time, at both companies.

If this is your cup of tea, it's worth reading about the Astor Place riots over Shakespeare performances in NYC

You are literally correct, but not engaging in the point. Meta shares in Scale are explicitly non-voting shares. Influence/control post minority investment matters as to whether the deal can be reviewed via Clayton Act. Meta did everything they could to avoid regulatory review with this transaction, and it worked.

This post doesn't quite comprehend why Meta made a 49% investment instead of an acquisition.

The path Meta chose avoided global regulatory review. FTC, DOJ, etc and their international counterparts could have chosen to review and block an outright acquisition. They have no authority to review a minority investment.

Scale shareholders received a comparable financial outcome to an acquisition, and also avoided the regulatory uncertainty that comes with govt review.

It was win/win, and there's a chance for the residual Scale company to continue to build a successful business, further rewarding shareholders (of which Meta is now the largest), which is just like wildcard upside and was never the point of the original deal.

iPhone 1 launched without cut/copy/paste

Socratic by Google still exists today and is a widely beloved app based on reviews. They had to rewrite the code and infrastructure, but "killing off the app" implies that they just shut things down. That never happened.

As for "happy story", I think the founders of Socratic learned a lot. Shreyans is just trying to share his learnings here. Not celebrate or mourn.

The deck plus the OEM dock with mouse, keyboard and monitor makes for a nice, cheap Linux desktop that you can grab-and-go to play games. The dock is really seamless, always works right away. If you're programming, I wouldn't recommend Deck without the dock.

echoes of United States v. Microsoft: https://en.wikipedia.org/wiki/United_States_v._Microsoft_Cor....

would be ironic if Mozilla played the next Netscape

Creation layer of the stack: Unity or Unreal.

This is a clever twist on unbundling. Well done.

Craigslist is misspelled (there is an "s" in it).

Giving me a photo credit would be courteous as the author of that image. https://thegongshow.tumblr.com/post/345941486/the-spawn-of-c...

You're talking about Presense, which is the point at which your brain flips and is convinced what its seeing is real.

Michael Abrash has a wonderful deck he wrote while at Valve about what it will take for VR to create Presense in the viewer. It's worth a read if you care about how we'll navigate past the uncanny valley of VR you describe: http://media.steampowered.com/apps/abrashblog/Abrash%20Dev%2...

Stephenson considers much of his position in "In The Beginning..." to be out of date at this point: http://en.wikipedia.org/wiki/In_the_Beginning..._Was_the_Com...

CSS (like all written works) is copyrightable, so if there is clear evidence that the CSS was copy/pasted, CB Insights would have a case on that basis. Damages would be negligible, so it would not be a case worth pursuing.

Does this BS marketing PR stunt qualify Nature-published research papers for Gates Foundation-backed research now? Meaning: does this announcement qualify Nature as "open" enough for the Gates Foundation?

Neal Stephenson identified this same issue, and created Project Hieroglyph as his solution. They're about to release their first anthology of optimistic, progressive sci-fi short stories this fall. http://en.wikipedia.org/wiki/Project_Hieroglyph

A good complement to this set of challenges is Dan Boneh's Crypto class on Coursera. The coursera class is more theory-driven, whereas these challenges are more practical... they mix well. https://www.coursera.org/course/crypto

I don't think the author jumps to conclusion prematurely. He hates all this factory, provider, blah, blah nonsense, and I can identify with the feeling.

He is however guilty of lacking originality in his post because Joel Spolsky made the author's point much better about 9 years ago: http://discuss.joelonsoftware.com/?joel.3.219431.12

I received a phishing email from the author. I guess he must have scraped my email address from a blog post I wrote about bitcoin and coinbase.

While I am glad he has made attempts to contact Coinbase, I felt like live execution of the attack was spammy, so my first instinct was the block the domain of the sender's email, which Coinbase passes through to me. In execution of his proof of concept, the author is likely badly ruining his spam score / sender score.

You need to read to the end of the post to understand PG's analogy. There is software that controls the charging rate of the battery that (if the claims are correct) could be fundamentally disruptive to batteries going forward, so the Altair analogy implies a potential software licensing strategy as an upside in this company.

Good summary of human being's advantage in the war. But don't forget bacteria's key advantage: iteration speed.

Bacteria may be blind and and random in their micro-level behavior. But the speed at which they replicate is multiple orders of magnitude faster than the speed at which we can test and iterate new defenses.

It's a race of blindingly fast random iterations VS top-down snail-paced logical defense. If human's are to win, I think they will need to bump up the defensive iteration time an order of magnitude or so.

Implementing the Trie (on paper, in C) was about 50% of my final exam on my intro to CS undergrad class at Stanford. We had never seen Tries befor that point, but we had done a variety of trees, so we had the necessary prior experience. It was a fun exam and a very memorable experience.

>> A doctor trained in India is no less qualified than a doctor trained in the US.

This might be objectively true (which you could reasonably determine through some combination of standardize testing and surgery complications/outcomes), but if you ask 100 people on the street in US whether they would prefer to see a US MD or an Indian MD for their heart surgery, I bet 97 or higher would prefer the US MD. The tiny remainder would probably show no preference.

Their is a brand power of a US medical education, which is especially strong for US patients.

Was thinking of getting one for the purpose of building a MAME cabinet. Does it run MAME out of the box?

This is a great page, but it assumes that I (the reader) am already certain I want to invest time learning Erlang. I think it would be valuable if the author would open the page with a brief paragraph on why it's worth investing the time to learn Erlang and some of its tradeoffs vs other languages.

The use of a phonebook analogy is clever, but I think the answerer should modify his answer to cover the O() running time of a trie data structure. A phonebook is essentially a trie. He talks about the "divide and conquor" approach to navigating a phonebook, but that's not how a computer would store the names. The way a computer navigates a phonebook would be interested in read about.

I have a dell xps 13 and the trackpad responsiveness is a disaster. I have tried endless combination of drivers (both Dell and third-party) and sunk 20 hours into debugging. I strongly recommend you do not purchase an XPS 13 if you do not plan to use it primarily with an external mouse.

Uninstall any extensions you have running in Chrome to get it feeling fresh again. In my personal experience (strictly subjective, not benchmarked), a clean chrome install runs faster than a clean Safari install.

I'm a VC, and when I read the line "buying sole control" I assumed that the Series A investment from Benchmark is intact, and their shares still vote on any protective provisions granted to the preferred stock, but Adam simply bought enough stock in the Series B (in combination with his originally granted founders shares) to do two things:

1) represent a majority of all outstanding stock (both common and preferred). 2) represent a majority of all preferred stock.

If the preferred stock votes on all protective provisions in an all-for-one manner, then Adam would control the preferred voting, assuming #2 is true above. All-for-one means that all classes of preferred stock vote together as one group, and each share, regardless of class, gives you one vote. If the governance issues that affect the preferred stock are not setup as "all-for-one" voting, then the other likely situation would be independent class voting, in which case Benchmark and Adam would each get their own veto (Benchmark would control the voting of the Srs A and Adam would control the voting of the Srs B, and so they could each vote the other party down). If Adam really has sole control, then preferred voting would need to happen in an "all-for-one" manner.

My explanation is just an assumption based on that one flip line, but that's how it reads to me.

Yes, this survey design is questionable by most standards.

But, it's not a survey design disaster and is likely a "directionally correct" reflection of internet users' actual behavior.

Are exactly 34.5% of the US Internet Population not using fb/twitter? Of course not, but I suspect that number is pretty close.

If you're building a social networking app, this is an opportunity.

I wonder if Marco hosts his own email? Email's the place where embarassing/sensitive information is most commonly stored, not the FS.