ealexhudson

The board of directors have no jurisdiction to investigate an "influence operation on their CEO", and even if they did, they have no power to do anything about it. Their jurisdiction is within the business.

If this was primarily an investigation of the external actors, this would be talking about what law enforcement are doing.

The wording of the announcement is pretty neutral but I think it's clear what is within and without their bailiwick. The title is a much more informative description than "Statement from the board".

Off your direct point, but I found this line in the article interesting: ""Jerry Solinas did provide the seeds to me sometime in Fall 1997."

The seed for P-224 uses a timestamp from 1998, and they also said the four binary curve seeds were timestamps from 1984 that were doubled - someone did srand48(time / 2) not srand48(time).

Did the P-224 seed get updated later / was not provided in 1997, or is there some other reason for this? Otherwise, it looks like one third of these values need additional excuses to get them to map to the time range in question...

In what ways do you expect a candidate for MP for Clacton to address the various calamities you mention?

Sadly another shot in the arms race that captchas started which just leads to increased inaccessibility.

It's interesting work for sure, but the end goal of separating out AI versus human consumers is tough. Indeed, if there was a lasting solution, that would be a substantial discovery that would quickly become very famous...

A lot of the solutions in the CPU space involve things like memory allocation flags, NX bits, canaries, etc. that fire deterministically. Those things are fundamentally not applicable to LLMs, and without those things modern software would be in a vastly worse place.

You could imagine that there are things to change around LLM architecture that will improve its ability to reject prompt "injection", but I think it's fundamentally true that from an information theory perspective there's no bright line between "instruction" and "input data" possible.

This plausibly demonstrates why a nonprofit may not be a great vehicle for some free software projects - while the nonprofit should do whats best for the project, if the main work is done by commercial sponsors then it’s crucial those sponsors feel the relationship is beneficial.

The reality is free software office apps require significant professional development input. Apache Open Office is the obvious example.

It’s a classic version of the tragedy of the commons. If Collabora goes off to its own thing, I struggle to believe they will maintain the development rate with new devs, and without development the TDF sponsorship will fall off.

I hope we are not looking back in two years time regretting this.

What would be great, and I don't know if @dang / the mods would take on requests like this, would be for bot participants to be allowed but the account flagged. So e.g. the user name just says "[bot] Zakodiac" or something.

As well as being an ethical approach - I think it's wrong to try to impersonate humans and/or not announce AI output as AI - it would also be handy for new filter options: all bot posts are OK, hide bot leaf comments, or hide all threads with bot comments. etc.

[edited as my robot unicode/emoji char didn't come through]

I think they needed to be clearer about what the actual requirement was.

If the requirement is, "Show the balance _as it was_ at that point in time", this system doesn't fulfil it. They even say so in the article: if something is wrong, throw away the state and re-run the events. That's necessarily different behaviour. To do this requirement, you actually have to audit every enquiry and say what you thought the result was, including the various errors/miscalculations.

If the requirement is, "Show the balance as it should have been at that point in time", then it's fine.

I suspect there is a bit of knee-jerk because so often this pattern is misapplied. I actually quite like the example in the article although I'm basically allergic to CQRS in general.

I think your point about write-ahead logging etc is a good one. If you need a decent transactional system, you're probably using a system with some kind of WAL. If you're event sourcing and putting events into something which already implements a WAL, you need to give your head a wobble - why is the same thing being implemented twice? There can be great reasons, but I've seen (a few times) people using a perfectly fine transactional DB of some kind to implement an event store, effectively throwing away all the guarantees of the system underneath.

Moore's law has kind of ended already though, and maybe has done for a few years, and even if you can make a chip which is faster there's a basic thermodynamics problem running it at full tilt for any meaningful period of time. I would have expected that to have impacted software development, and I don't think it particularly has, and there's also no obvious gain in e.g. compilers or other optimization which would have countered the effect.

I don't want to sound too dismissive, but all these arguments have been brought up time and again. The move from assembler to high level languages. The introduction of OOP. Component architecture / COM / CORBA / etc. The development of the web browser. The introduction of Java.

2018 isn't "the start of the decline", it's just another data point on a line that leads from, y'know, Elite 8-bit on a single tape in a few Kb through to MS Flight Simulator 2020 on a suite of several DVDs. If you plot the line it's probably still curving up and I'm not clear at which point (if ever) it would start bending the other way.

Surely the example can be "obvious" because it's simple/clear. I don't think they're commenting on whether _finding_ the example is obvious...

Sure, but would we really want to tell liquidators to manage assets for best eventual return rather than just convert everything to cash? In this instance, in hindsight, sure - you'd want the other thing, you want the bitcoin not the cash. But this feels like the exception that proves the rule.

The trustee's reports on FTX's internal processes were damning. Even they had held their Anthropic on the way up, who's to say their internal FTT ledger and black holes in the Alameda books would not have eclipsed that?

The issue wasn't that crypto markets in general were down at that point; the issue was they were doing frauds.

The reason is simple: it allows you to do the install using "sync" in all cases, whether the lockfile exists or not.

Where the lockfile doesn't exist, it creates it from whatever current is, and the lockfile then gets thrown away later. So it's equivalent to what you're saying, it just avoids having two completely separate install paths. I think it's the correct approach.

Perhaps your content quality meter needs a recalibration?

Agreed. "Dessert" vs "desert" - mistaking these two is often not a grammatical error (they're both nouns), but is a spelling error (they have quite different meanings, and the person who wrote the word simply spelled it wrongly).

The Judges impose the punishment set out in the law; they don't make this stuff up.

The alternative is Judges letting people off just because they're politicians. That seems like an extremely poor precedent to set, those in political life should be held to higher standards.

Within UK dialect there would be some significant differences in many of these words, even ignoring the meddle/mettle examples - farrow/pharaoh is easily distinguishable, too.

I would say, though, that to people _outside_ the dialect, there may be many more words that are indistinguishable. Listening to Scots speakers requires a lot more effort for me because to my ears, many of the differences in the words are extremely subtle.

If we think of the 'budget' as being similar to a bandwidth limit on video playback, there's a kind of line below which the picture starts being pretty unintelligible, but for the most part that's a slider: the less the budget, the slightly less accurate playback you get.

But because this is clean data, I wonder if there's basically a big gap here: the codec that encodes the "correct rule" can achieve a step-change lower bandwidth requirement than similar-looking solutions. The most elegant ruleset - at least in this set of puzzles - always compresses markedly better. And so you can kind of brute-force the correct rule by trying lots of encoding strategies, and just identify which one gets you that step-change compression benefit.

I think you're right about the essential ingredient in this finding, but I feel like this is a pretty ARC-AGI specific result.

Each puzzle is kind of a similar format, and the data that changes in the puzzle is almost precisely that needed to deduce the rule. By reducing the amount of information needed to describe the rule, you almost have to reduce your codec to what the rule itself is doing - to minimise the information loss.

I feel like if there was more noise or arbitrary data in each puzzle, this technique would not work. Clearly there's a point at which that gets difficult - the puzzle should not be "working out where the puzzle is" - but this only works because each example is just pure information with respect to the puzzle itself.

No, it's incorrect and/or badly worded. The author is right that a machine cannot author things, and the stuff that the LLM might create de novo would not have copyright protection. But it's missing the point when the argument is that existing authored works could be generated via an LLM, and the authorship/copyright is already established.

I'm not sure that's completely true.

Having read MS code and starting to generate new code that is heavily inspired - sure, that's not copyright infringement. But, if you had memorized a bunch of code (and this is within human capability; people can recite many works of literature of varying length with total accuracy, given sufficient study) - that would be copyright infringement once the code was a non-trivial amount. The test in copyright is whether the copying is literal, not how the copying was done/did it pass through a human brain.

This scenario rarely comes up because humans are, generally, an awful medium for accurate repetition. However, it's not really been shown than LLMs are not: in fact, CoPilot claims (at least in its Enterprise agreements) to check its output _does not_ parrot existing code identically. The specific commitment they made in their blog post is/was, "We have incorporated filters and other technologies that are designed to reduce the likelihood that Copilots return infringing content". To be clear, they only propose to reduce the possibility, not remove it.

LLMs rely on a form of lossy compression which can sometimes give back verbatim content. I think it's pretty clear and unarguable that this is a copyright infringement.

We don't really know what GPT-4 "is". I remember reading a number of relatively well-informed suggestions that there are a number of a models inside there, and the API being interacted with is some form of outer-loop around them.

I don't think the location of the outer-loop or the design of it really makes much difference. There is no flock of birds without the individuals, the flock itself doesn't really exist as a tangible thing, but what arises out of the collective adjustments between all these individuals gives rise to a flock. Similarly, we may find groups of LLMs and various outer control loops give rise to an emergent phenomena much greater than the sum of their parts.

I think to be clear, brute force generally means an iterative search of a solution space. I don't think that's what this system is doing, and it's not like it's following some search path and returning as early as possible.

It's similar that a lot of wrong answers are being thrown up, but I think this is more like a probabilistic system which is being pruned than a walk of the solution space. It's much smarter, but not as smart as we would like.

The point of doing the interviews is to ground the product in actual feedback from (potentially) paying customers.

This is a make-believe facsimile of that. This is what insights and feedback might look like, but I can't imagine that it would be usable as such...

It's not even that: it's a device for non-dermatologists, to decide if they need to make a referral. "Cleared" is definitely the right language here, this might only be a Class 1 medical device, it's not making any diagnostic claims.

It's a bit chicken and egg though: I would expect that a grant would be much more likely to be granted for an event that had a decent amount of groundwork already done? For new events especially, you'd want to see some commitment.

Imagine a politician photoed reacting to a member of the public in disgust, except the face was stitched in from moments after. Or, worse, someone captured at the scene of a bomb reacting before the bomb went off?

We have an inbuilt set of assumptions about causality that this AI now violates. That's potentially huge in some very specific scenarios...

By "independent" I don't mean "functions separately". Of course the CEO sits on the board. Sometimes the CFO is on the board too, although subordinate executives usually _should not be_ (they may _attend_ the board, but that's a different thing).

But fundamentally, the CEO _reports to_ the board. That's the relationship. And in a 501(c)3 specifically, the board have a clear requirement to ensure the company is running in alignment with its stated charter.

Whether or not this board got that task right, I don't know, it doesn't seem likely (at least, in hindsight). But this type of board specifically is there for oversight of the CEO, that's precisely their role.