It's the age of automation. They likely got more than one courtesy automated reminder e-mails. If one doesn't respect automated reminders, one will respect automatic service cut-out.
I've done a lot of C++ with GPT-4, GPT-4 Turbo and Claude 3.5 Sonnet, and at no point - not once - has any of them ever hallucinated a language feature for me. Hallucinating APIs of obscure libraries? Sure[0]. Occasionally using a not-yet-available feature of the standard library? Ditto, sometimes, usually with the obvious cases[1]. Writing code in old-school C++? Happened a few times. But I have never seen it invent a language feature for C++.
Might be an issue of prompting?
From day one, I've been using LLMs through API and alternate frontend that lets me configure system prompts. The experience described above came from rather simple prompts[2], but I always made sure to specify the language version in the prompt. Like this one (which I grabbed from my old Emacs config):
"You are a senior C++ software developer, you design and develop complex software systems using C++ programming language, and provide technical leadership to other software developers. Always double-check your replies for correctness. Unless stated otherwise, assume C++17 standard is current, and you can make use of all C++17 features. Reply concisely, and if providing code examples, wrap them in Markdown code block markers."
It's as simple as it gets, and it didn't fail me.
EDIT:
Of course I had other, more task-specific prompts, like one for helping with GTest/GMock code; that was a tough one - for some reason LLMs loved to hallucinate on the testing framework for me. The one prompt I was happiest with was my "Emergency C++17 Build Tool Hologram" - creating an "agent" I could copy-paste output of MSBuild or GCC or GDB into, and get back a list of problems and steps to fix them, free of all the noise.
On that note, I had mixed results with Aider for C++ and JavaScript, and I still feel like it's a problem with prompting - too generic and arguably poisons the context with few-shot learning examples that use code that is not in the language my project is.
--
[0] - Though in LLMs' defense, the hallucinated results usually looked like what the API should have been, i.e. effectively suggesting how to properly wrap the API to make it more friendly. Which is good development practice and a useful way to go about solving problems: write the solution using non-existing helpers that are convenient for you, and afterwards, implement the helpers.
[1] - Like std::map<K,T>::contains() - which is an obvious API for such container, that's typically available and named such in any other language or library, and yet only got introduced to C++ in C++20.
[2] - I do them differently today, thanks to experience. For one, I never ask the model to be concise anymore - LLMs think in tokens, so I don't want to starve them. If I want a fixed format, it's better to just tell the model to put it at the end, and then skim through everything above. This is more-less the idea that "thinking models" automate these days anyway.
We won't be eliminating anything until we give up on only ever working directly on plaintext single source of truth code. AI is automating tedium that's otherwise impossible to automate because we're stuck in a 1970s Unix paradigm and can't let go.
I have a feeling it's eventually going to be possible to work around this by setting up live AI upscaling to 4k - should be a year or two before that starts to work in real-time on higher-end gamer GPU.
(Of course I suspect DRM will then get extended deeper into the GPU.)
HDCP already encrypts everything starting from video decoding to the signals being sent to the monitor. It really can't get "deeper" than that, other than maybe fixing whatever current vulnerabilities pirates are using to bypass it.
Then it'll get re-extended to 1080p again? Point being, I can see live 1080p to 4k upscale via generative AI becoming a feasible alternative to watching the original 4k stream where DRM prevents you from accessing the latter.
> Presumably you're trying to trigger ketogenesis.
Only if you're actually on a keto diet; most weight loss routines run a calorie deficit without triggering ketogenesis.
Also there's this whole thing about set points - you probably got to your weight somehow, and if it wasn't through consistently making stupid choices over many years, chances are this is what the body learned to consider an equilibrium state. Which means that, if you start running a calorie deficit, it's going to fight you every step of the way. It will happily scale down performance to conserve energy instead of burning the accumulated fat, so you'll just be slow and groggy but not lose weight. There's been reported cases where people got mental illness-level obsessive thoughts about food, which appeared when they were hungry, and stopped when they ate enough.
The degree of this problem varies between people, but it's generally not that easy to effectively lose weight, and some people simply lost the genetic/environmental lottery on this.
> The degree of this problem varies between people, but it's generally not that easy to effectively lose weight, and some people simply lost the genetic/environmental lottery on this.
Yea, I think your eating and food habits you learn as a child and teenager tend to shape you for life. I was malnourished as a child (I was considered a picky eater) but as an adult I've found it quite easy to keep within the caloric bounds a doctor told me to keep to. One thing I've noticed is that hunger just doesn't bother me the way it does with people who struggle to lose weight or with binge eating—mostly a pain in the ass (I need to remind and force myself to eat), but occasionally something I'm grateful for as I watch the people around me struggle on an existential level with their cravings and bodies.
I also don't have a sweet tooth, and I put that on not being allowed sweets as a child except under very exceptional circumstances. I'm also a (thankfully recovering) alcoholic, so don't mistake this for being generally good at avoiding cravings.
> Also there's this whole thing about set points - you probably got to your weight somehow, and if it wasn't through consistently making stupid choices over many years, chances are this is what the body learned to consider an equilibrium state.
The idea of set points is controversial and not necessarily the accepted scientific consensus.
> People who don't need to be chronically medicated should not be.
If they're like typical westerners, they already self-medicate with coffee every day, with alcohol occasionally, and a big fraction of them (though much less than couple decades ago) also treat themselves with tobacco smoke - and ironically, weight loss is one of the few benefits some people actually use to defend their smoking.
Do they need all that medication? Well, it's socially unfavorable to say so wrt. alcohol, but ask any of the daily coffee drinkers whether they need their morning coffee...
So to pay for the war to stop a new Soviet Union, a bloc of countries that all fear their alliance will turn into a new Soviet Union, must become a new Soviet Union?
I would say it's even better: if flushes out the engineers who never bother to grasp what "science" actually referred to. you'd think empirical observation and iteration would be an easy to grasp concept, but there are individuals that people get tripped up by something unidentifiable. you can spot a mile away. You can see it here, too. Not sure what causes the mental block but it seems to form serious social impediments.
reply