ChickenOverlord
No bio...
User ID: 218
ChickenOverlord
18hr ago
The funniest part to me is that Bob Dole actually outlived John McCain, even though Dole's age was a major angle of attack when he ran against Clinton in 1996.
ChickenOverlord
1d ago
The thing is, I think a lot of SMEs and domain experts don't realize just how bad things are even without AI.
I do, I've written about it here (and elsewhere) even before AI became a big thing. My grand unifying theory of modern software development is roughly:
-
There is greater demand for software than there are competent developers/engineers capable of delivering it
-
There are "enough" incompetent software developers to meet the demand
-
Companies hire more incompetent devs than competent because there are far more competent than incompetent
-
Companies create retarded processes and systems like scrum and agile because carefully managing and babying your incompetent devs is the only way to eke out something functional
-
The competent devs get sucked into these shitty processes (because they want/need jobs, and companies are terrible at identifying talent so they get lumped in with theretards), so even they generally aren't making quality software because of the bullshit processes etc. they have to deal with
ChickenOverlord
1d ago
Absolutely, either way I think OP is jumping to judgement prematurely until we know the specifics of the deal
ChickenOverlord
1d ago
Where is the actual 10 point proposal? Or is that being kept confidential while it's still under negotiation?
EDIt:
Assurances of "safe" passage through the strait of Hormuz, but no assurance of "free" passage.
The "COMPLETE" part implies free. If it isn't what actually happens feel free to correct me.
ChickenOverlord
1d ago
OpenBSD still finds a buffer overflow every year or two. It's definitely better than 95% of big software projects out there, but it isn't perfect. Definitely not trying to minimize what Mythos actually found though.
ChickenOverlord
1d ago
I'm not skeptical about every single aspect of AI, my main skepticism is over its ability to build and maintain complex systems (usually in the form of codebases that are more than a basic bitch CRUD app). Finding vulnerabilities is definitely something I've always thought was within the capabilities of AI, my biggest concern is the signal to noise ratio. So I'm curious how many false positives Mythos found that they had to filter through to find the 4 examples they list as ones it actually found.
ChickenOverlord
1d ago
If that's the biggest red flag you can find?
I mean I'm sure I could find others if I tried.
(And I suppose I should thank you for listening to others when they asked you to try repeating your recent experiment with Opus instead of Sonnet. That makes you a better skeptic than many I have the displeasure of knowing on this forum.)
Thanks, I try.
Plus, they've already said they're not going to make Mythos public, even if some of the benefits will trickle down to the next Opus. That is not something a company that is desperate for money or willing to ignore safety would do.
They've only said the preview of Mythos won't be public, the final release will be.
ChickenOverlord
1d ago
Biggest red flag to me that this is more marketing puffery overselling capabilities than reality:
Early indications in the training of Claude Mythos Preview suggested that the model was likely to have very strong general capabilities. We were sufficiently concerned about the potential risks of such a model that, for the first time, we arranged a 24-hour period of internal alignment review (discussed in the alignment assessment) before deploying an early version of the model for widespread internal use. This was in order to gain assurance against the model causing damage when interacting with internal infrastructure.
I.e. "This AI could be utterly devastating even if we only let it loose on our internal network. We'd better be super duper extra careful and cautious before we let it loose. 24 hours ought to be fine, what could we possibly miss in such a massive time window?"
Surely these are just average joes and average janes? Do you mean to tell me if the woman trains for a couple years, and is healthy / responsive to training, she wouldn't be stronger than the majority of men that don't train? Don't most women just avoid actual strength training / bulking out of temperament/desire for their body to look a certain way and not out of inability to do it and see results?
Doesn't seem to be the case, as the repeated instances of 14 year old high school boys beating (and usually outright demolishing) women's Olympic teams, World Cup teams, etc. would demonstrate:
US Women's soccer team loses to an under-15 boy's team, score 5-2
Australian Women's soccer team loses 7-0 to an under-16 boy's team
High school boy's team beats Olympic US women's team at hockey, 2-1
https://www.espn.com/olympics/news/story?id=2281644
And there are countless more instances of this. I'm sure these high school boys are more fit than typical boys their age, but male physical strength generally peaks between 25 and 35, so they are likely physically weaker than the median untrained adult male.
The numbers I've seen over the years (I'd have to try and track them down) are that a woman has to be in roughly the top 5 to 10% of women to beat a male in the bottom 5% of male strength.
Edit: found a chart of female vs male grip strength, you can see that they barely overlap. Grip strength is hardly the only form of strength, but from my recollection other forms of strength show similar trends:
ChickenOverlord
1d ago
Suppose you are interviewing for a job, and I come in in clothes which I have been wearing for a week which have tomato sauce on them. For most positions, this would instantly disqualify me, and rightly so.
If I'm hiring someone to reverse engineer the firmware of a competitor's product, I'm hiring whoever is the most competent for the job, even if it means hiring the sexually deranged catboy wearing programming socks. And for something like reverse engineering firmware, I'd venture a guess that somewhere in the range of 50 to 75 percent of the qualified candidates are catboys (or aesthetically similar).
The same goes for fixing the rot in Western Civilization. The overwhelming majority of candidates capable of fixing it are going to share a lot of Trump's bull in a china shop aesthetics, it's just kind of the nature of the sorts of people capable of what is needed. Sure, some candidate with will and ability to get things done and with the aesthetics of JFK might exist out there, but I'm not going to vote for Democrats that will keep deepening the rot in the meantime while I wait.
ChickenOverlord
1d ago
My position - that threats of genocide are more concerning you have the ability to carry them out - is not complicated or ambiguous
Then you should have stated that clearly at the start of this? Instead of vaguely saying that when Iran says death to America it's different.
so what is the point of asking "So is it ok to threaten genocidal destruction so long as you don't have the capacity to actually carry it out?"
To find out what your actual position is, since you didn't clearly state it.
Do you actually think I believe that?
No, hence my asking if that's what you believe.
ChickenOverlord
1d ago
The problem is that nowhere in my post do I say or even imply that Iran's rhetoric is acceptable.
But you do say that it's different than Trump's genocidal threats, with Iran's capacity to act on what they said being the only difference you mention. I'm really not sure what you're going for here bro.
ChickenOverlord
1d ago
I don't know how you inferred that from what I wrote
You were replying to someone who asked if you raised similar complaints over Iran calling for death to America for decades with the argument that a big difference between the two is that the US is currently bombing Iran and has the capacity to inflict significant damage while Iran currently isn't and cannot. I'm not saying anything excuses anything else here, I'm just replying to the argument you made.
ChickenOverlord
1d ago
So is it ok to threaten genocidal destruction so long as you don't have the capacity to actually carry it out?
ChickenOverlord
2d ago
If they're going to enshittify the AIs, it'll have to happen after one company gets sufficient market dominance that swapping to a different one isn't trivial.
Microsoft already had Copilot start inserting ads into pull requests without consent a week or two ago, I think that counts as enshittification.
ChickenOverlord
2d ago
if America blew up a primary school in my country I'd start chanting Death to the Great Satan as well
They've been chanting that particular line for decades now though
ChickenOverlord
2d ago
I'll try that later this week when I get a chance. Maybe next time I'm stuck in an awful meeting for two hours
I'm sure there are more, but these immediately come to mind. There are four of us trying to make these things work, and we all keep running into the same problems again and again. It's not just me - even people with dramatically different writing styles and thought processes are seeing the same thing. I feel like I'm taking crazy pills, because a lot of people I know in real life are experiencing the same pain, but on the Internet it seems like I'm a huge outlier.
My most competent co-worker, a Russian guy who got his start writing assembly back in the 80's, was the most enthusiastic about/interested in AI person that I knew. He was always trying out the latest models from OpenAI, Google, and Anthropic. He was also running his own local LLMs and diffusion models locally. He even dropped $4-5k on a DGX Spark late last year. And even he seems to be getting disillusioned/losing interest in AI, he doesn't seem to think it's going to be able to achieve anything remotely close to the promises and hype. Though I will note that the push from our upper management to use AI hasn't pleased him much either, especially since the project we've been working on for the past year (modernizing a giant mess created by our Indian coworkers. They weren't using package management at all, they were literally emailing around zip files full of DLLs for years, I got pulled into 4 hour long calls to fix dependency conflict issues in prod once every 2 or 3 months) was very much not aided by AI, but management insisted we find a way to use AI on the project regardless.
ChickenOverlord
2d ago
this video is great and captures my frustration with a lot of the software developers who I have to work with
Sadly the parody developer in that video would be more competent than most of my Indian coworkers, so I wouldn't be surprised if AI could replace them. But instead it will be one of the slightly more competent Indians generating mountains of barely functional AI slop (instead of the small hills of slop my coworkers currently generate).
I feel Americans are far too quick to congratulate themselves on the topic of freedoms and rights.
Not at all. Our track record is far from perfect, but we still somehow manage to completely eclipse every other country on earth when it comes to speech rights, in spite of our failures and shortcomings. We can call our politicians idiots without getting arrested [1], and in the rare cases when cops have overreached for that sort of thing the courts have shut it down.
1: https://www.dw.com/en/germany-greens-habeck-presses-charges-over-online-insult/a-70793557
Alright, AI bros, follow-up from last week. I was able to secure access to Claude Opus 4.6 at my job, and I gave it the same prompt that I had given to Sonnet. It overlooked the authentication part of the HTTP client library completely this time in what it generated. In a follow-up I asked it to extract out the common logic for the authentication portions specifically. It didn't do that, instead it generated a class with two helper methods.
The first helper method was just a thin wrapper around System.Text.Json for deserializing the response. There's an optional flag to pass in for when case insensitive deserialization is needed, and nothing else.
The second helper method was something for actually making the HTTP calls. The strangest part with this one is that it has two delegates as parameters, one for deserializing successful responses, the other for handling (but not deserializing) error responses. It didn't do anything to split out handling of the 2 different ways to authenticate at all.
The issues with what was generated (for both the API client as a whole, and for the authentication part of the code specifically) are numerous, here are a small handful that I identified:
-
It assumes that an HTTP 200 code is the only successful response code, even though some endpoints return 202, 207, and more.
-
It assumes that all endpoints return plaintext or JSON content, even though several return binary data, CSV data, etc.
-
It didn't do null checking in several places. I assume it was mostly trained on C# code that either didn't do null checks correctly, and/or on code that doesn't use the nullable reference type feature that was added in C# 8 (back in 2019). Regardless, the null checks are missing/wrong regardless of whether nullable reference types are enabled or disabled. Also it always checks nulls with == or != null. This works 99% of the time, but best practice is to use "is null" and "is not null" for the rare cases where the equality operator is overloaded. Once again, I assume this is because most of the training data uses == and !=.
-
It doesn't handle url query parameters (nor path parameters), it assumes everything is going to use a JSON body for the request.
-
It uses the wrong logging templates for several of the logging calls. For example, the logs for an error response use the log template for logging the requests that are sent. Even more troubling is that it removed all the logic for stripping user secrets out of these logs.
There are quite a few more issues, but overall my experience with Opus was even worse than my experience with Sonnet, if anything. AI bros still in shambles. I definitely have zero fears that AI will replace me, though I'm still definitely fearful that retarded C-suite execs will think it can replace me.
My post from last week about using Claude Sonnet: https://www.themotte.org/post/3654/culture-war-roundup-for-the-week/426666?context=8#context
Edit: Just saw a very relevant post over on Orange Reddit about this very topic: https://news.ycombinator.com/item?id=47660925
ChickenOverlord
3d ago
Or it just outsources the pollution to China. Not saying to pollution from making solar panels is equivalent to the pollution from burning coal, but clean energy isn't quite as clean as many of its proponents like to portray it.
Was hard to find any hard numbers (most sources just talked about how much coal plants emitted, but not how much the tech to limit emissions costs per kwh). I did find this though: https://about.bnef.com/insights/industry-and-buildings/us-coal-plants-face-new-rule-capture-co2-or-shutter/
ChickenOverlord
3d ago
You also have to account for restrictions and regulations on coal and other sources. If coal and oil plants were allowed to burn as dirty as their 18th century equivalents, there would be no beating them on cost per kwh. Whether or not that would be a good thing is a separate question, but the comparison isn't a completely fair one because even without subsidies it's not a completely level playing field.
Hegseth was just counteracting decades of systemic racism and sexism against white men in the military.
More options
Context Copy link