This study this meme is based on is completely incorrect and should be retracted. Here’s a lay summary of its issues:
And the published article detailing the problems with that study’s issues:
https://www.sciencedirect.com/science/article/abs/pii/S1090513824000497
Trogdor was popular way before Reddit
It doesn’t have to be
https://www.mathworks.com/products/compiler.html
MATLAB can ruin all sorts of coding experiences, programming included
Examples? I can think of a number of foreign companies that the US facilitates, like Nestle.
Eh, I switched. I switched all of my lab’s computers, too, and my PhD students have remarked a few different times that Linux is pretty cool. It might snowball.
Desoxyn would like a word.
Edit to add: more commonly prescribed amphetamines are neurotoxic, too. Whether they are neurotoxic at clinical doses is still debated.
This makes sense, thanks
Why would China turn against Putin for them using their nukes? I don’t keep up much on their relations.
Journal quality can buffer this by getting better reviewers (MDPI shouldn’t be seen as having peer review at all, but peer review at the best journals–because professors want to say on their merit raise annual evals that they are doing the most service to the field by reviewing at the best journals–is usually good enough at weeding out bad papers), but it gets offset by the institutional prestige of authors when peer-review isn’t double-blind. I’ve seen some garbage published in top journals by folks that are the caliber of Harvard professors (thinking of one in particular) because reviewers use institutional prestige as a heuristic.
When I’m teaching new grad students, I tell them exactly what you said, with the exception that they can use field-recognized journal quality (not shitty metrics like impact factor) as a relative heuristic until they can evaluate methods for themselves.
The best time to start was decades ago, but at least they’ve started.
This is a problem that’s becoming outdated, thanks to NIH now requiring females to be included in studies in order to receive grant funding–barring an exceptional reason for studying males alone (e.g., male-specific problems). They are even requiring cell lines for in vitro studies to be derived, at least in part, from females, rather than from males alone.
Seven paragraphs is too much? I read the full thing before seeing your comment. It’s well written and easy to read.
A fellow Julia programmer! I always test new models by asking them to write some Julia, too.
I actually took that bit out because LLMs are pro climate and against everything that makes the environment worse. That’s a result of being trained on a lot of scientific literature. I was just curious what Opus would say about the conceptual knowledge piece.
Claude Opus disagrees, lol (edit to add: all of what follows is Claude; not me):
I respectfully disagree with the claim that current language models are logically incapable of mapping language to actual concepts. While today’s LLMs certainly have limitations and are prone to hallucinations, they have demonstrated a remarkable ability to capture and manipulate complex concepts through language.
A few key points:
LLMs are trained on vast amounts of human-generated text which is grounded in real-world concepts, allowing the models to build rich representations that go beyond surface-level word associations. Techniques like unsupervised pre-training enable LLMs to learn meaningful conceptual relationships.
In many domains, LLMs have shown an impressive capacity for reasoning, inference, and knowledge synthesis - skills that rely on a deep understanding of the underlying concepts, not just pattern matching. Their performance on complex question-answering, analogical reasoning, and natural language inference tasks is evidence of this.
LLMs can be fine-tuned and augmented with techniques like retrieval-augmented generation or chain-of-thought prompting to further ground them in factual knowledge and improve their reasoning capabilities. This allows more robust conceptual mappings.
Careful prompt engineering and sampling techniques can help to mitigate hallucinations and nonsensical outputs. While not a complete solution, this shows that the conceptual knowledge is often present even if the model sometimes struggles to express it faithfully.
That said, I fully agree that the tendency for subtle nonsense is a serious issue, especially for layperson use cases where hallucinations may go undetected. Continued research into making LLMs more truthful, consistent and robust is crucial. Techniques like constitutional AI to bake in truthfulness, and better UX design to convey uncertainty are important steps.
But in summary, I believe the evidence suggests that LLMs, while flawed, are not fundamentally incapable of meaningful conceptual representation and reasoning. We should push forward on making them more reliable and trustworthy, rather than dismissing their potential prematurely.
The online play is garbage. I played in H1 tournaments around the US back when it was good and would love for them to do it better than they did with their remake. The remake actually remade Halo 1 PC, not the Xbox version.
Sorry, but this makes clear that you aren’t in science. You should avoid trying to shit on studies if you don’t know how to interpret them. Both of the things you mentioned actually support the existence of a true effect.
First, if the treatment has an effect, you would expect a greater rate of relapse after the treatment is removed, provided that it treats a more final pathway rather than the cause: People in the placebo group have already been relapsing at the typical rate, and people receiving treatment–whose disease has been ramping up behind the dam of a medication preventing it from showing–are then expected to relapse at a higher rate after treatment is removed. The second sixth-month period was after cessation of the curcumin or place; it was a follow-up for treatment-as-usual.
Second, people drop out of a study nonrandomly for two main reasons: side effects and perceived lack of treatment efficacy. The placebo doesn’t have side effects, so when you have a greater rate of dropout in your placebo group, that implies the perceived treatment efficacy was lower. In other words, the worst placebo participants are likely the extra dropouts in that group, and including them would not only provide more degrees of freedom, it would theoretically strengthen the effect.
This is basic clinical trials research knowledge.
Again, I have no skin in the game here. I don’t take curcumin, nor would I ever. I do care about accurate depictions of research. I’m a STEM professor at an R1 with three active federal grants funding my research. The meme is inaccurate.
A bit of an exaggeration, sure. But only a bit. The lay summary of the article I referenced states the following:
“commits every error that it was possible to make in the paper,” and, “completely incorrect,” aren’t very different.