davemp

Eugenics is a taboo because there is a tempting trap to over simplify and make assertions that are not actually supported by the data.

We know that IQ is hereditary to an unknown degree. We have some evidence that IQ relates to intelligence.

We don’t know really know which genetic variables are responsible. Even simple features like height are thought to have ~12k variables.

We don’t even really have a good definition of intelligence (see any AGI comment chain).

I would say that assuming we have good enough data or models to base important decisions on is unscientific.

Decisions like improving education and nutrition don’t really need to help.

I always have a problem when folks bring up idiocracy because the of the eugenics angle. It’s extremely unlikely that people are getting inherently stupider, just less educated. The former is some sort of prophecy of doom and the latter is actually actionable.

Easy to spot in a contrived example is not:

> impossible to express in Rust

I’m not going to argue with Rust folks who misrepresent the language.

    let foo = [1, 2, 3];
    unsafe {
        *foo.get_unchecked_mut(4) = 5;
    }

Not sure why Rust evangelists always seem to ignore that unsafe exists.

I appreciate the detailed reply and that there’s subtlety here.

I read the linked Bartz case. It’s disappointing that it seems limited to only the copying of books into a data set and not the result of training LLM on protected works. This is not the “use” that I was discussing and not very interesting.

The plaintiffs didn’t even challenge that the outputs of the LLMs infringe. They judge seems to agree (at least by omission) that fair use wouldn’t apply but that the outputs were transformative and in cases where they aren’t:

> [anthropic] placed additional software between the user and the underlying LLM to ensure that no infringing output ever reached the users.

So this is not true:

> he [the judge] details exactly why the facts of the case and prior case law find that training an AI on copyrighted material is fair use

The plaintiffs also make really awful arguments about “memorizing” and “learning” that falsely anthropomorphize LLMs. Which the judge shoots down.

If we’re going to give LLMs the same rights as humans, there’s unlikely to much of an argument.

I think there’s potential for an argument about how LLMs use “compressed” versions of protected works to _mechanically_ traverse language space. It would be subtle and technical so maybe not likely to work in our current context.

Yet you still wasted your own time and everyone else’s time with a reply that has even less substance.

I was making an argument based on quotes from the actual legal code and you’re saying pions who don’t use the exact correct terminology shouldn’t even consider what should or shouldn’t be legal? What a load of junk. This is a democracy. We’re supposed to be engaging with it.

I replied to someone saying that it’s fair use, which presupposes that it’s a derivative work.

Claiming LLMs are fair use is ridiculous bordering on ignorant or disingenuous.

Here’s the 4 part test from 17 U.S.C. § 107:

1. the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

Fail. The use is to make trillions of dollars and be maximally disruptive.

2. the nature of the copyrighted work;

Fail. In many cases at least, the copy written code is commercial or otherwise supports livelihoods; and is the result much high skill labor with the express stipulation for reciprocity.

3. the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and

Fail. They use all of it.

4. the effect of the use upon the potential market for or value of the copyrighted work.

Fail to the extreme. There is already measurable decline in these markets. The leaders explicitly state that they want to put knowledge workers out of business.

- - -

Hell, LLMs don’t even pass the sniff test.

The only reason this stuff is being entertained is some combination of the prisoner’s dilemma and more classic greed.

> No. Reading something, learning from it, then writing something similar, is legal; and more importantly, it is moral.

Machines aren’t human. Don’t anthropomorphize them. The same morals and laws don’t apply.

Considering we can only approximate irrational numbers, I’m not sure that’s a given. Maybe we’ll have a breakthrough with some type of analog computing, but we could also just hit physical limits on energy or precision.

So often these AI articles mis or ignore the Test Oracle Problem. Generating correct tests is at least as hard as generating the correct answers (often harder).

I’m actually starting to get annoyed about how much material is getting spread around about software analysis / formal methods by folks ignorant about the basics of the field.

There’s like 10 replies telling me it’s not actually that complex then describe an equally or more complex solution that’s probably missing some of the security or features I have. I guess I wrote “complicated to set up” when I meant “complicated stack”.

A heavyweight blob of python running in a container talking to other containers running node-js on a Linux VM on a Linux host behind layers of networking to toggle 1 bit of information over a UHF link is really complex even if it only took 1 button click. My solution just removes the linux VM.

I’m not even saying I struggled, it took me like 1-4hrs depending on if you include the non HA stuff. I can clone the repo I made, populate the secrets/gateway, and type ‘make’ to rebuild my setup on a new machine. Everything is “zero-trust” with signed SSL certs, dynamic DNS+wireguard to tunnel in from a stable url, and room for more services. It’s cool, but it’s only toggled 1 light for years.

I’m not giving untrusted devices unfettered access to my lan and an airgapped network sounds more complicated tbh. VLANs aren’t really that bad with good networking gear.

I use an RPi 4 w/ 4GB of RAM and a 1TB usb 3.2 SSD and it flies through all my workloads. Though I don’t do any video encoding things.

Fun project. Though it’s kind of unreal how complicated it is to set up HA and I literally do this for a living, both embedded sw and backend web dev.

Docker compose with a zwave management server, reverse proxies for TLS, vlan isolation for the server, macvlan for HA container so it does see the host network, etc, etc. All to turn on and off a lightbulb with the sun. All the while AI is telling me to configure things insecurely.

I think when I get some more spare time, I’d like to write a statically linked program that handles a zwave controller and basic automation scripting. No IP networking needed for my lightbulbs. Then it wouldn’t feel risky to just make a system user and udev rule to give it permissions to the controller, and run with systemd.

Professionally, sending our code off prem is not an option. Frankly I don’t understand why executives are okay with AI companies training LLMs on their IP. Unless they own a significant stake in the AI company I guess.

Personally, it’s been decent for generating tedious boilerplate. Though I’m not sure if reading the docs and just writing things myself would have been faster when it comes time to debug. I’m pretty fast at code editing with vim at this point. I’m also hesitant to feedback any fixes to the AI companies.

I’ve found “better google” to be a much more comfortable if not faster way to use the tools. Give me the information, I’ll build an understanding and see the big picture much better.

Do you walk around with rotating face coverings and disguise your gait?

The onus is on the organizations spying on the public and the Govt. for letting them do it.

> The current project I'm working on has about 3 copies of every component because nobody bothers to clear up after themselves

In these cases you know the duplication and dead code is there because of sloppiness and Chesterton’s fence doesn’t apply.

> https://en.wikipedia.org/wiki/G._K._Chesterton#Chesterton's_...

Ignorance of why something exists is not a good enough reason to destroy it.

Leave parenting to the parents. Stopping bad/evil parents is not possible.

The tiktok/youtube recommendation algorithms will undoubtedly cause more harm to minors than wandering onto an adult website and learning about how babies are made.

Extremely debatable. Especially because there is no "The Turing Test" [0] only a game and a few instances were described by Turing. I recommend reading the original paper before making bold claims about it. The bar for the interrogator has certainly be raised, but considering:

- the prevalence "How many |r|'s are in the word 'strawberry'?" esque questions that cause(d) LLMs to stumble

- context window issues

It would be naive to claim that there does not exist, or even that it would be difficult to construct/train, an interrogator that could reliably distinguish between an LLM and human chat instance.

[0]: https://archive.computerhistory.org/projects/chess/related_m...

Amazon hedging $35 billion on an IPO (we know the odds of AGI with LLMs are vanishing small) is concerning. The hedge signals that they think it's a good "greater fool" investment, otherwise they should be happy to take equity in a company they think will be stable/profitable.

Does it even need to be explicitly stated? Closed linux userlands.

I’m not going to write a detailed response to something that’s extremely close to what an LLM responds to “what UB does zig have?”

Arguing about whether certain static analysis should be opt in or opt out is just extremely uninteresting. It’s not like folks are auditing the unsafe blocks in their dependencies anyways.

If you want to talk about actual type system issues that’s more interesting.

This is pretty close to saying Rust is not very different than C because it has the unsafe keyword. That is, either an ignorant (of Zig) or disingenuous statement.

> IMHO the zig footgun story with respect to UB behavior is largely unchanged relative to C/C++

The only major UB from C that zig doesn’t address is use after free afaik. How is that largely unchanged???

Just having an actual strong type system w/o the “billion dollar mistake” is a large change.

Brute forcing a problem with a perfect test oracle and a really good heuristic (how many c compilers are in the training data) is not enough to justify the hype imo.

Yes this is cool. I actually have worked on a similar project with a slightly worse test oracle and would gladly never have to do that sort of work again. Just tedious unfulfilling work. Though we caught issues with both the specifications/test oracle when doing the work. Also many of the team members learned and are now SMEs for related systems.

Is this evidence that knowledge work is dead or AGI is coming? Absolutely not. I think you’d be pretty ignorant with respect to the field to suggest such a thing.

Very good news. Happy to see Zig getting more investment. I already use the build system for production systems and am looking forward to writing some more Zig when projects allow.

Having a modern language that attempts to directly model low level systems is very important.

Windows prioritize phoning home and data collection over UX. If you have a corporate install you’ll also have negligent EDP software killing your boot times.

You can get fast boot times on linux if you care to tweak things.

I’ve always just blamed the extreme bloat of the web and lack of design around poor connections for 2bar lack of performance. HN usual works fine on but that’s about it for sites I visit.