davrosthedalek

But: "After posting this on Hacker News and recieving [sic] almost no substantive input, discussion, or response on the actual content of the article, I decided to rewrite all of the prose in my own voice. If anyone complains about my verbosity or sentence structure — as they usually do, which is the reason I originally let the AI write the prose, among other reasons obsoleted by templating — they can go fuck themselves."

So rewritten in his own voice. Maybe the m-dashes are from GLM, maybe from the author.

Haven't looked at the code, but the allocated memory could be larger than necessary to make "off-by-one" or "off-by-a-few" errors less deadly. Then zeroing it out makes it even less so. Defense in depth.

Or it's an allocation for an arena? The zeroing might help trigger 0 derefs earlier if the overrun happens for the object that are then allocated in the arena (and not by allocating more objects than the arena can provide)

Many statistics were presented. In the view of the author (and I think he is correct), none of them show evidence for an increased bug rate from Claude. That is absence of evidence (...for the increased bug rate).

The two examples you bring are not claims of absence of evidence, but claims of evidence of absence. The author takes the result as evidence that there is no effect. As I wrote, the author shouldn't do that, because indeed you cannot distinguish between "no effect exists" and "no effect observed". But again, these are (wrong) claims for evidence of absence.

The author can absolutely claim: I did these statistical tests, and none showed evidence that there is an effect. Absence of evidence. It's not a claim that there will never be evidence. Just that there is none from these tests.

Edit: To convert the absence of evidence into evidence for absence, indeed you need to understand the statistical power of your test, and how it is affected by alternate hypotheses. And for that, without having done the math, having only two data points seems very thin.

Yes, I do for example.

And the author discussed the use of AI pretty exhaustively in point 0 of the post.

No. It's a description of the result of the maybe underpowered study. the underpowered study did not find evidence. Evidence is absent. Because it is underpowered, it's not evidence that the effect is absent.

The claim is not "two experimental conditions did not differ". The claim is "The data do not show evidence that the experimental conditions did differ".

He can't make the evidence of absence claim, but he can absolutely make the absence of evidence claim.

The point is that you should look only at the first smlen entries, which would be 0 for this case.

That's how it used to be. At least at my alma mater, it changed a lot when they switched from Diploma to Master/Bachelor

The problem here is that it will work for now, but how do you make sure the LLM talks to the student, and not a different LLM? I guess vision models FTW?

In the end, will be build a GAN loop?

Why am I now reminded of corewars?

Absolutely university has to change. But it's not a simple change. I say this as a professor for Physics:

My colleagues say "We must fully embrace AI as a tool". I agree. But how do you teach it? It's a moving target, and you can't even give homework like: "Research <this topic> with an LLM of your choice, and submit the transcript" because they can do that, or they can just copy the task into an LLM and have the LLM do it. It becomes meta quite quickly.

And independent what and how we teach, we have to change how we assess a students learning result:

The first thing we have to change is that homework needs to be completely ungraded. Reviewed and corrected, yes, but not part of the grade. That's the only way to make sure that people who don't want to cheat have to cheat anyway to compete with those that do.

Second, all exams have to be in person. Online, cheating is so trivial it's not even funny (many students are so stupid about it that we have a pretty clear idea what's going on). In person, we have maybe 2-3 years until we have to make sure its proctored and people's glasses are checked. I think in less than 10 years, local mobile AI will be good enough so even a Faraday cage will not help.

Maybe we have to go to oral tests only.

Of course, none of this scales. Some of our intro courses have a thousand students.

Any ideas are much appreciated.

On top of that, the flow of electrons is seriously slow compared to the speed of the current.

I am not sure, but I think you might misunderstand what emergent means. Take chaotic systems, in the math sense. Chaos is a well defined property of, say, iterated dynamic systems.

A linear dynamic, say x_n+1=lambda x_n, or x_n+1=(1-x_n), is never chaotic. But if you multiply them, x_n+1=lambda x_n (1-x_n), it depends on lambda if the system is chaotic.

None of the components are chaotic. But for specific combinations, chaos emerges as an property.

In physics, the mass of mesons and the nucleon is emergent. It's completely different from the constituents' mass. Different from an atom, where its mass is very close to the mass of its nucleus and its electrons.

I mostly agree, but I see two points that might be problematic:

a) The brain might have an entropy source (then it can't be modeled as a function). Trivially to fix, and in some sense, with diffusion models starting from random numbers, AI has done so.

b) The hidden function might be not computable. I would have no idea how that would work, but I think this is what it boils down to if people say "the human brain is more than a machine".

But it could be. It's just less efficient.

Obligatory link to mars.com: https://chaos.if.uj.edu.pl/~wojtek/MARS.COM/

Yes, and I think a free-version user might produce more support requests than a commercial user for two reasons: 1) commercial/professional users might feel more entitled to support, but typically have a better understanding of linux and more versed in fixing stuff themselves. -- and more importantly -- 2) They probably have a dedicated setup where they can run the AMD-blessed distro

Indeed, time synchronization across detectors is always tricky. Distributed clocks get messy at ATLAS dimensions. WR allows to distribute pretty good time sync over large detector systems. Sometimes still not good enough though. Time-of-flight detectors try to get to single-digit ps level, and almost by definition, you have to synchronize two detectors that are some distance apart.

Twice the path delay + the time it takes to send the return packet. I assume WR does this in hardware to get a predictable time.

Or, maybe it's beginning of the end of social media. Might not be the worst idea.

This. It's extremely frustrating, AI can be a 24/7 tutor, but too many students use it to do their work instead.

We have to really rethink how and what we teach, and how we evaluate. Scoring (non-handwritten) homework is pointless, even contra-productive (because it incentivizes cheating, even for the students who don't want to, just to not be outscored by the cheaters). Hand-written homework means the students at least have to have read the work once...

And soon, with AI glasses, even in-classroom tests will be difficult.

I agree that tables are not great, but figures? Seems to work pretty well, better than word for sure.

Yes, and dark matter will interact with visible matter gravitationally. When we say "doesn't collide with normal matter", it is not exact. It just means "the interaction length is very very long". Same for neutrinos. Their interaction length is huge as well: 1 TeV Neutrinos have an interaction length of 2.5 million kilometers.

We measure the matter distribution by its affect on light (strong/weak lensing). We also measure the matter distribution by the amount of light coming from it. The results are not the same. The simplest explanation is that there is matter which does not produce or reflect light via e/m, i.e. it is dark. Dark Matter.

We know of particles which behave the same way. Neutrinos for example.

If there would be a distribution of dark matter that explains the poltergeist, we could measure that distribution of dark matter.

We can measure the mass distribution on astronomical scales. We "see" the dark matter. Just not with light.

The "pass through matter" is a consequence of not interacting electromagnetically. That's not that uncommon. Think neutrinos. (Also, "not" might just mean "very much reduced")

It might just be a reflection of the architecture the universe simulation is running on...

The correct response is not to press any button, but to find whoever set up this game, and remove their capability to harm anyone.

What title?

Taylor series have a quite different convergence behavior than a general polynomial approximation. Or polynomial fit for that matter. Many papers were written which confuse this.

For example, 1/(x+2) has a pole at x=-2. The Taylor series around 0 will thus not converge for |x|>2. A polynomial approximation for, say, a range 0<x<L, will for all L.

Both Western Digital and Kioxia have reported their 2026 Flash/Hard drive production capacity is sold out.

Micron killed Crucial to focus on AI.