TruePath

Yes, it changes the nature of the work. Back when you started coding there were people experiencing the same thing about shifting to higher level languages. What some of them liked was the efficency and aesthetic of using just the right assembly language trick and good compilers with high quality instruction selection took that away. I'm sure there were programmers who missed the days of punching in hex into memory.

We start to understand those old fogeys who we blew past we were young once we get to their age. It's the way of the world.

If the AI is good enough to truly implement the whole thing to a similar level of reliability without copying it then who cares. At that point you should be able to decompile any program you want and find enough information inside that an AI can go write a similar quality program from the vague information about the call graph. We've transcended copyright in computer code.

If it can't and it costs a bunch of money to clean it up then same as always.

OTOH if what is actually happening is just that it is rewording the existing code so it looks different then it is still going to run afoul of copyright. You can't just rewrite harry potter with different words.

Note that even with Google vs oracle it was important they didn't need the actual code just the headers to get the function calls were enough. Yes it's true that the clean room isn't required but when you have an AI and you can show that it can't do it a second time without looking at the source (not just function declarations) that's pretty strong evidence.

That's assuming you won't need anyone to manage the purchased software platform and its integrations and that you need a full time engineer to maintain your version.

Could you clarify exactly what you think is an illegal tie-in? Because it seems like what you are upset about is literally the opposite -- Anthropic unbundling their offerings so you aren't required to buy the ability to offer third party access when you purchase the ability to use Claude code and their other models. Unless I really misunderstand you, your complaint is literally thaf

The laws prohibiting tie-ins don't make it illegal to sell two products that work well together. That's literally what the laws are designed to make you do -- seperate products into seperate pieces. The problem tie-in laws were designed to combat was situations like Microsoft making a popular OS then making a mediocre spreadsheet program and pushing the cost of that spreadsheet program into the cost of buying the OS. That way consumers would go "well it's expensive but I get excel with it so it's ok" and even if someone else made a slightly better spreadsheet they didn't have the chance to convince users because they had to buy it all as one package.

Anthropic would be doing something much closer to that if they did what you wanted. They'd be saying: hey we have this neat Claude code thing you all want to use but you can't buy that without also purchasing third party access. Now some company offering a cheaper/better third party usage product doesn't get the chance to convince you because anthropic forced you to buy that just to get claude code.

Ultimately this change unbundled products the opposite of a tie-in. What is upsetting about it is that it no longer feels to you like you are getting a good deal because you now have to fork over a bunch more cash to keep getting what you want. But that's not illegal, that's just not offering good value for money.

I guess I don't even really understand the objection. That's how ALL mathematics works. You specify some axioms or a construction and then reason about objects that satisfy those constraints. Some of them like the complex numbers turn out to be particularly useful.

But it's not fundamentally any different than what we do with the natural numbers. Those just feel more familiar to you.

Whoever wrote this just doesn't understand who apple's main customers really are. Yes, devs may be a high impact customer base but most of apple's customers are people like my mom who struggles with the difference between Gmail the app, Gmail the web page and Gmail in apple mail and is reasonably worried about scams and viruses because she knows she isn't really tech savvy enough to spot them. If she is going to run AI on her apple products it can't be 'well it probably won't delete your data.'. It needs to be something she can be sure is safe and is limited to the access she gives it.

That's a really tough problem. I'm not even sure yet google can pull it off.

What is the usecase for using ssh at all where you don't need to be resistant against timing analysis? Either it's not sensitive and you can use telnet (if necessary after using ssh to authenticate) or the game (or other stuff on the connection) might be sensitive and you need traffic analysis resistance.

If you get clever and write a client to ensure sensitive data like passwords or email are sent in a burst you could just use an encryption library just for that data instead.

Lots of the real world vulnerabilities out there exist exactly because of people choosing to support a range of crypto algorithms.

Sure, if it's an internal tool you can recompile both ends and force a universal update. But anything else and you need to stay compatible with clients and anytime you allow negotiation of the cryptosuit you open yourself up to quite a few subtle attacks. Not saying that choice about go is clearly a good one but i don't think it's obviously wrong.

To clarify the point in the other reply -- imagine it sent one packet per keystroke. Now anyone sitting on the network gets a rough measurement of the delay between your keystrokes. If you are entering a password for something (perhaps not the initial auth) it can guess how many characters it is and turns out there are some systemic patterns in how that relates to the keys pressed -- eg letters typed with the same finger have longer delays between them. Given the redundancy in most text and especially structured input that's a serious security threat.

I couldn't think of something worse than demanding Amazon decide what is counterfeit or violates regulations and policing that rule. The law on both those points is far too complex for the result to be anything but Amazon blocks what the big brands tell them to and protect them from competition.

Amazon is essentially a logistics company with a search engine. It doesn't really make sense to have them enforce regulations or counterfeiting rules than it would to make UPS and google. It's not like they hide who the seller is on any item (it's listed as sold by).

What your complaining about is a fundamental consequence of anything that lowers the barriers to selling goods. You once needed to buy a storefront to sell retail goods, later you at least needed sufficient name recognition for people to visit your website -- that investment gave anyone whose goods were counterfeit as well as regulators assets to seize.

But just like making it easy for every citizen to publish their thoughts means we see lots of hate and dumb shit online -- anything that lowers the barriers to selling retail goods (in general a good thing) will make it easy to sell counterfeit or defective crap.

In the long run, I suspect tech will make reputable 3rd party evaluations easier to access but let's not blame Amazon for not becoming an arm of the state and judging what is and isn't legal.

That gets a bit tricky in terms of what you mean by valid programs. I presume what you mean is that you can't write a compiler that accepts every function which always returns the borrowed reference and reject every piece of code which fails to do so.

Though it's technically a bit different than the halting problem as this issue remains even if you assume that the function terminates -- you only want to show that the reference is returned assuming the code terminates if it isn't returned because it enters an infinite loop that's not a leak.

There is no inherent benefit in going and expressing that fact in a type. There are two potential concerns:

1) You think this state is impossible but you've made a mistake. In this case you want to make the problem as simple to reason about as possible. Sometimes types can help but other times it adds complexity when you need to force it to fit with the type system.

People get too enamored with the fact that immutable objects or certain kinds of types are easier to reason about other things being equal and miss the fact that the same logic can be expressed in any Turing complete language so these tools only result in a net reduction in complexity if they are a good conceptual match to the problem domain.

2) You are genuinely worried about the compiler or CPU not honoring it's theoretical guarantees -- in this case rewriting it only helps if you trust the code compiling those cases more for some reason.

It is worth remembering that before artificial lighting the normal sleep pattern is to sleep in two segments and get up in the middle to have sex, talk, do some chores, change the baby etc. And I believe a normal sleep cycle is something like 90 minutes so if you *really* sleep in 3 hour chunks that's actually probably healthier than the normal us sleep schedule. I mean many of us find the usual wake at 7am alarm schedule to be pretty damn brutal. However, if what you actually do is kinda lie around check your phone in bed etc and then get blasted awake by an alarm that's a different story.

Some of the happiest and most satisfied people I know are academics who work 80+ hours a week because they love what they do. You don't need to sacrifice health to work more than 50 hours a week. And realistically there is no long term health damage from what he is doing. And yes you can trade money for better health, though in the first world you rapidly hit diminishing returns but if your sleep deprivation can make the difference between being able to afford good health care or moving to an area with less pollution it absolutely can trade off like that.

I really enjoyed the video but I think it can be a bit misleading. It gives the impression that it is the universal approximation property that make neural nets so effective when, of course, the learning algorithm: memorize the training data and on input x return the output y which was associated with the input nearest x -- also has the universal approximation property. Given training data from a continuous function f sampled via a distribution with non-zero density on interval I it will a converge to f (uniformly if I is a finite, and hence compact, interval). Nor does the geometric explanation at the start have anything to do with why neural nets are so effective.

I'm honestly unsure of the theoretical reason neural nets are so effective at language processing. But it certainly requires some characterization of the problem space (e.g. problems where the function to be learned has such and such property) since there are plenty of mathematical techniques for approximating well-behaved continuous functions with certain features that way outperform neural nets at learning various classes of mathematically nice functions..

This is a problem that can only be fixed by legislation. Other things being equal DRM that's convenient for the vast majority increases book sales and consumers aren't willing to pay enough more for DRM free books to cover the difference -- much less understand the difference between 'owning' books and 'if we feel like it' licenses. And even if a general (i.e. not O'REILLY) ebook/ereader vender just wanted to do the right thing it's far from clear you could write a license that is flexible enough to allow pivoting with changing market conditions, get good deals with publishers but not so flexible as to be meaningless. And it's hard to sell a more expensive DRM free option alongside since, as the DRM free price goes up each sale is more likely to be sold to a 'pirate'.

The government can invoke the Sherman antitrust act and other laws. YOU can't (at least not generally).

Maybe I'm missing something, but seems pretty brittle to me. From what I can tell, since the decryption key is entirely derived from the information in the PSP header (SA/SPI) a bad actor who observes PSP encrypted packets from some other sender can simply copy that inner packet and the target machine will decrypt it as if it came from that sender.

The document acknowledges this but basically leaves it to other aspects of the network stack to defend against this (maybe there is some extra protection provided by the ICV check). Google's stack seems carefully designed to be secure in this way but it feels brittle.

Wouldn't it have been better to require a checksum of some of the exterior headers (source IP??) inside the encrypted section to block attempts to repackage the same encrypted content inside another packet. Or is that somewhere in there and I'm missing it?

I'd like to see a charity that retrospectively sends people like this cash. Creating both the sense that you'll be financially and socially rewarded for this kind of brave action would do a lot to make the world better.

Damn't I didn't mean compact connected. I meant that if you have a convex region you can extend it. Obviously, non-convex regions can raise problems (e.g. maybe tropical rainforest can't transition to tundra in the space of a single tile). But it should work with convex regions.

I don't get the worry about the NP complete problem of tiling a finite area.

Yes, of course, if you allow arbitrary adjacency rules then tiling is going to be NP complete but the goal here is to generate realistic tilings then it's reasonable to assume that if you have a compact connected region that obeys the tiling rules and any square there is some tile that is allowed to be placed in that square.

I mean this is how the real world works. You don't have holes in reality so if it's possible to have a part of the world that looks like such-and-such there must be a valid description of the stuff next to it.

Which is exactly why you wouldn't waste time signing a useless quote.

Just sign a message attesting that YOUR NAME is satoshi. No more work. Done once.

Umm, he obviously does. Money, fame and influence over future bitcoin decisions.

Imagine your Satoshi and you need cash for some reason. Reveal your identity and immediately get a bunch of offers for book deals and speaking engagements. You run into career problems, reveal who you are and you can probably snag a position as a Google fellow.

More concerned about your legacy or some crypto-anarachist vision? Revealing yourself could help you influence both the public policy debate on encryption and bitcoin's future (speaking engagements offer more influence than some signed messages on a mailing list).

Finally, what about credit for your contribution. Being known to your friends and family.

Sure, he may still want to remain anonymous but to suggest there isn't the temptation to go public is silly

Umm why? For the most part a professor is a talking text book that stands in front of the class. You can go download charismatic people with fancy approaches to the material online. However, I don't think this really makes much of a difference (choosing the right curriculum does though).

The one place I've seen professors have real impact is in their ability to both answer innovative and novel questions and convey a sense of excitement and interest in a subject. While these only affect a small fraction of any given class they are the important value add and they are what someone who is personally passionate about the field and deeply knowledgeable about it brings to the table.

Worse, what seems like "good teaching" can often reflect a counterproductive attitude to the subject.

Everyone I know who has become a mathematician has done so DESPITE (not because) of their pre-college teachers (some have had the good fortune to run into one or two exceptions but still struggled past teachers who attempted to ruin the subject). These teachers were chosen entirely for their teaching ability but to no avail since without the perspective of a research career they weren't teaching people how to duplicate a computer algebra system instead of understanding the math. Focusing on accurate symbol manipulation and memorization is in active tension with understanding but is an easier rote to solving the kind of problems you can ask on AP tests.

College students are big boys and girls. They can read the damn textbook if needed. I spent an entire semester at caltech (my most productive one) without attending a single class (I had my friends turn in my homework in section). However, having professors available who could accurately convey (through course content, discussion and even research projects) what a discipline was about was invaluable.

Let's give up the charade people and let professors play videos of better speakers in lecture but lend their expertise to answering questions and talking to students.

Should departments have standards they expect their lecturers to enforce?

People would actively welcome new approaches to teaching math. It's fucking frustrating not getting through to your students or having them dislike the course. Most of the faculty has tenure and the focus is largely (but not entirely) on research. THE IDEA THAT PEOPLE ARE THREATENED BY HIS SUCCESS IS ABSURD.

He is succeeding in the truly obvious way. Lowering standards and making it easy for students to ace his class without the skills the university has deemed necessary.

If he was really so deeply gifted at teaching he could have just used the same final exam as another faculty member and taught his students that material. Why doesn't he?

While I agree that the most important thing is convincing the 1 person out of 1,000 to be a math major (or physics or whatever) I'm afraid the university isn't ok with handing everyone else As without mastering a certain minimum skill set.

Besides, the math department isn't a free agent in this. Other departments expect students who have passed math 1A to have certain skills. If the math department was free to teach math 1A as an advertisement for mathematics (the way other departments get to teach their base courses) it would be a much different class. Hell, it wouldn't even be about calculus.

So long as the physicists, economists, chemists etc.. etc.. say "we need students with skills X, Y and Z" there has to be a class whose content (and hence a decent grade in) is X, Y and Z. The name of that class is 1A and it helps no one to just teach a different class under that name without reforming the system.

The math department ALREADY OFFERS OTHER CLASSES DESIGNED TO DO EXPOSE STUDENTS TO MATH IN EXACTLY THE WAY YOU MENTION. These are courses that deliberately avoid the equations and rote memorization that generated so much fear in high school and are designed to be low stress while showing off the wonder and creativity of mathematics.

Unfortunately, few people ever take these courses because students don't want any more math than they have to take and the guidance counselors and university won't push or require them.

As you said about the final (which I wholeheartedly agree with) if you give students the choice how is it the department's fault you take it. Students have a choice to approach math in a more exploratory, more encouraging fashion but they choose to do the minimum needed to take whatever other class they need. It's ironic because that's also what makes calculus so hard for them...they want to apply a rule and confidently move on...not think hard about a math problem in ways that might be dead ends.

That is what's so sad about math education. Understanding comes from trying to fit the concepts together and even (especially) professional mathematicians fail 90% of the time. Students, especially those who haven't done well in math find that very distressing. Hell, pick any subject you aren't good at (which has clear success/failure conditions) and really try to do it well. It makes you feel dumb and you give up. Add to that years of pre-college math (often taught by teachers who themselves seek the reassurance of rote procedure) which brain washes the curious kids into believing that doing math is just applying procedures and it's an uphill battle.

Ultimately, I think the problem is that we try and force kids who aren't really interested to know math they don't need. Just like other subjects let the students who want to learn come to you. With the advent of modern computer algebra system no one benefits making kids memorize rote procedures beyond basic arithmetic since without understanding they will never apply it in the real world.

Teaching positions for grad students aren't predatory. They are really a form of financial assistance. In return for teaching 3 hours a week 2 hours sitting in your office with the occasional student visitor (with say 5 hours of prep work that could be done during office hours and 3 12 hour days grading exams). That's 8 hours of work a week which works out to a pretty decent hourly salary.

It's hardly luxury and when the state/university system yanks funding out during your tenure THAT is somewhat predatory but usually most departments find a way to tighten belts and make sure current students have sufficent funding.

Hmm, my experience in 2001-2008 wasn't the same. While there were certainly funding issues the department worked pretty hard to try and do what it could to help. Then again technically I was in the logic group so things were a bit different for me but I'm conveying what my math grad officemates and friends experienced (i.e. they said they had pretty much the same situation).

Yes, some people didn't finish but math is hard and while the department is large and can be impersonal there were no barriers and efforts were made to assist the students.

The cheap labor argument doesn't really hold water (at least during my time). More students wanted teaching positions than positions were available.

Having actually taught (well TAed)in the UC Berkely math department I can say that generations of both graduate students (who have most of the contact with students) and professors have struggled to the best of their ability to teach calculus in an engaging and interesting way to their students. Now research talent and teaching talent don't always come together but there were plenty of dedicated and talented teachers. Dr. Coward is basically claiming that he is so amazingly talented (or that not one of the multitude of intro calc teachers over the years was competent) that he single handedly does the impossible and rises head and shoulders above all these other teachers.

Even though he quotes from the letter from the math department disturbingly selectively I think he gives away what is really going on in that quote:

"They also reveal some significant differences between your practices and what has been typical in our department [...] I hope and expect that you will be able to align more with our standards for the remainder of this semester and during the next academic year."

Surprise surprise, the students like the guy who doesn't make them learn (those in 16A/B aren't continuing on in a technical field but only take the course to satisfy requirements and/or because their counselors stress the importance of having math on your transcript).

If this isn't suggestive enough consider his statistics: his classes have 400+ students enrolled but the supposedly equivalent class taught by a faculty member only has 100. If he isn't cherry picking an 8am lecture (in which case we should be even more distrustful) this difference is so drastic as to be extremely suspicious. My experience TAing these very courses was that students were quite willing to put up with a bad lecturer (or poor TA in my case) to have a nicer schedule (leave earlier, avoid trudging back and forth etc..) but they were very cogniscent of even minor differences in supposed ease of grading and subject matter.

Besides, anyone who rights an open letter like this while supposedly appealing a deciscion is an ass who would be a nightmare to work with. Even if you (unwisely) believe he was unjustly terminated delibrately impugning the motives and integrity of the people he supposedly wishes to remain colleagues with is just a super asshole move.

I've met great teachers who were very dedicated to improving mathematics education. Every one of them avoided self-aggrandizement and did their best to share their approaches and experience with other educators. The mathematics department at UC Berkeley would be ecstatic if someone figured out how to teach intro calc content in a way that invigorated the students. Hell, it's almost a holy grail their as generation after generation of grad students (and professors) burn themselves out against students whose experiences in highschool have created an instinctive dislike of mathematics and the unshakable belief that math is about formulas and procedures (meaning they instinctively panic if a problem requires creativity...which is sadly what makes math interesting and fun).

---

Look, I have big issues with the way UC Berkeley (and most other universities) teach intro calculus courses. It's the 21st century and no one is served by forcing uninterested students to learn computational techniques by rote any computer mathematics system can easily accomplish. Either you are going to go into a technical discipline where having a conceptually grasp of integrals matters or you can just blindly plug them into a computer.

So give up making students memorize the various tests for series convergence, tricks for integrating etc.. Hell give up teaching computations at all in 16A/B. Slow it way the hell down and just do basic qualitative reasoning about functions, derivatives and integrals, e.g., focus on things like if we know the derivative is positive here what does that tell us about the function.

However, that choice to abandon the established curriculum is rightly something that needs to be handled by the department as a whole. Indeed, I suspect it would have long since been adopted if other courses (like intro econ...yet again not really for econ majors) didn't assume students had these tools. Changing the math curriculum in this way would first require other subjects change theirs...something that won't go over well and would take time.