I mean it might be a waste of money, but I think part of what they're trying to accomplish is not what's good generally, but also how certain people may respond to different foods, which they hope to leverage for doctors to personalize diet recommendations:
“There can be this public perception out there that ‘Oh, everyone knows what you’re supposed to eat, but it doesn’t work for me.’ But if this is for me and based on me,” people might be more likely to follow the plan, Nicastro said.
I wouldn't trust any powdered turmeric. Especially in the quantities people are consuming it these days.
Buy it fresh and grind it yourself. It tastes better that way anyway.
I've seen a scientific researcher do literature review manually.
Search, refine, print, read, highlight, collate.
Lots of time can be saved by automating those steps (and many researchers don't enjoy it so their job satisfaction could be increased). Also the resulting output could be improved if the researcher had a well structured summary to use as the foundation of their outline.
Improve search with semantic search (search by concept not keyword)
Improve refinement by preprocessing and summarizing
Don't print, display clean and concise data.
Summarize, cite and display.
This stops short of literature based discovery, you have to bring your own research question.
We've also had some luck finding a gap in existing research.
We did a POC where we scraped pubmed and graphed study results by concept. We then used the graphed concepts to explore the conceptual space.
It seems that vitamin D protects against cancer and heart disease. It seems that vitamin D supplementation protects against cancer but not heart disease. Is this because of some previously unknown effect of sun exposure (the primary natural source of vitamin D) or is it just that people with adequate vitamin D go outside a lot more and therefore also get more exercise? Don't know, would love to read the paper if someone studies it ;- )
It's actually possible to use LLMs to assist literature reviews. We built a product that works smashingly. The key is to keyword extract, use a vector database and do search based generation.
Our key insight is that the process of citation needs to be handled outside the LLM. They're good for text processing and summarization but as you said, the LLM itself is poor at citation.
> Our key insight is that the process of citation needs to be handled outside the LLM.
I can imagine taking the citation tree and using the LLM to compact the hypotheses, results, etc. for each node in the tree and sticking that in the vector database could get you pretty far.
I founded a company working to help companies use AI to organize their private knowledge.
We have focused on semantic search and knowledge graphs, but we started integrating a chatbot last week and it seems a short leap from where we are.
We'd be happy to help implement something.
You'll certainly want an embedding database.
The open models are getting pretty good, but you'll want to stand up a testing framework. I have a reasonably good model running on a desktop machine in my office with a reasonably priced consumer grade nvidia GPU.
We also have some tactics and practices around hallucination prevention that we'd be happy to share.
Feel free to reach out:
human at summitlabs.ai
Wasn't that the previous wave of remote work? And didn't it end badly for IT and software engineering?
I'm currently cleaning up a mess left by some well-meaning but unskilled and improperly managed overseas teams. They spent months and didn't deliver. My team cleaned it up in a few weeks but it wasn't cheap. And consider months of opportunity cost for lost execution time. In a startup with a fixed runway that can mean death.
The IT story is even worse. Poor outcomes, unhappy employees, data breaches. I heard from a neteng friend that when Google purchased Motorola the offshore neteng contractor was so bad Google that classified the Motorola network as actively hostile. The contractor was so deeply entrenched it was impossible to get them out and they were actively serving malware. Any machine that visited the motorola network had to be wiped before being allowed back on the corp network.
So yeah, if that's the next phase we already know how it ends.
They'd act like a solar sail. Putting them in L1 would not keep them positioned. It would be possible to design an orbital scheme that would keep them in place, but it wouldn't be simple. As other people have mentioned, this idea seems impractical and secondary effects haven't really been considered. Glad to hear someone's thinking creatively though.
Here's a simpler plan:
1) Crack down on the deliberate and aggressive misinformation pretending climate change isn't real. We allow this under the guise of free speech, but it's obvious it's being perpetrated to deliberately push a known falsehood. We know who's doing it, we know why. We know where the funding originates and where it goes. There's no ethical dilemma, no slippery slope. Just one lie that's killing the planet.
2) Shift all subsidies from carbon intensive energy to renewables and make it permanent. (This accelerates decarbonization and costs literally nothing).
3) Create a replacement for the Paris agreement but this time focused around collaboration, technology sharing, and a stack-ranked collaborative plan. (Develop and share technology around methane reduction, grid stabilization, electrification etc)
That's it. 3 step plan, net zero cost. Guaranteed to work. Won't ever happen if we don't address the public opinion campaign turning us against each other. But it is way easier than space bubbles.
The problem is not intractable. It's quite solvable. The one blocker is that we're working against ourselves. Let's stop working against ourselves and actually work together to save the livability of our planet.
The US will never be a leader in climate change reduction till we address tiny the minority that inexplicably wishes to profit from the destruction of the only place in the universe that is perfect for us. This is the literal garden of eden, we are willfully destroying it.
No plan can succeed if we don't address the cause of resistance to the plan. That's deliberate deception and manipulation of public opinion by a handful of billionaire oil barons.
The word "neglect" as it is used creates an incorrect framing. But is actually a great word. The whole problem with Nuclear in the US is that we neglect the projects. Our plan around large, long-term nuclear creates a vicious incentive structure. Recouping billion-dollar investments leads to to nuclear plants being run as long and as cheaply as possible.
This is a recipe for disaster.
Until we figure that part out, we are correct to hold off building them.
Perhaps a better framing is "We've neglected to figure out an incentive system that allows us to build, safely maintain and safely decommission nuclear projects."
More accurately, "There is no scenario in which building nukes produces economically competitive power, so there is no value in a capacity to build them." We will need capacity to decommission them for quite some time to come. Each one decommissioned is a $billion not available to spend adding 10+ GW of renewables, but patching them up would cost even more.
Can someone tl;dr? Not that the paper is too long, but I think I'd need to go back to university and major in biology or statistics to understand what they discovered and more importantly what it might mean.
When you have TC (total cholesterol) < 200 mg/dL (milligrams per deciliter) - each 1mmol/L (millimoles per liter) reduces mortality.
When you have TC > 200 mg/dL - each 1mmol/L increases mortality.
The confidence interval was high.
> In the age groups of 18–34, 35–44, 45–54, 55–64, 65–74, and 75–99 years, each 1 mmol/L higher TC increased mortality by 14%, 13%, 8%, 7%, 6%, and 3%, respectively (P < 0.001 for each age group), for TC ≥ 200 mg/dL, while the corresponding TC changes decreased mortality by 13%, 27%, 34%, 31%, 20%, and 13%, respectively, in the range < 200 mg/dL (P < 0.001 for each age group). TC had U-curve associations with mortality in each age-sex group. TC levels associated with lowest mortality were 210–249 mg/dL, except for men aged 18–34 years (180–219 mg/dL) and women aged 18–34 years (160–199 mg/dL) and 35–44 years (180–219 mg/dL). The inverse associations for TC < 200 mg/dL were stronger than the positive associations in the upper range.
Speculation in the discussion section is that while lower TC (below the limits cited above) is correlated with lower cardiac risk, it seems to be correlated with a higher risk of death. So the implication is there's something else bad for your health about being out of the sweet spot they observed in the data.
Imagine you have container ships named HDL and it is loaded with containers called triglycerides. You don't know how many containers you have. You just know you have this many container ships. Do you have too many containers? Do you have too many ships? That question cannot be answered with just one absolute number.
This is why the HDL to triglycerides ratio exists.
Implies a causal relationship. In reality, it's much more likely that causality points the other way, and an underlying disease like cancer is causing low cholesterol: https://www.ahajournals.org/doi/10.1161/01.cir.92.9.2396
Is it correct that 230 mg/dl is 5.6mmol/l? We've always been told that we must keep the number below 5 at all costs for health, but this seems to be as bad as being above 6.3. Am I reading this correctly? If so, current medical advice is way off.
TLDR: this observational study replicated the known phenomenon that (seemingly paradoxically) low cholesterol is correlated with increased risk of mortality. It's an ongoing debate whether low cholesterol is directly bad for you, or it could itself be a marker of underlying disease like cancer. The authors admit this study is unable to shed further light on that debate due to lack of data on specific cause of death.
I also find this unreadable but for a different reason. I'm on mobile and it's rendering unreadably small. On the one hand maybe I should put more time into my setup. On the other hand maybe we, the target audience of this sort of content should comment when the format is unusable. If your power users find mailing list popups annoying to the point they go elsewhere don't you want to know? Isn't a comment on hacker news a great way for someone to learn what their readers like or don't like?
> Please don't complain about website formatting, back-button breakage, and similar annoyances. They're too common to be interesting. Exception: when the author is present. Then friendly feedback might be helpful.
If the author isn't actually present (which, as far as I can tell, they are not) it just clutters up the comments. There's no actual discussion happening here, just a lot of "wow i sure don't like this thing [website] does" and that doesn't provide very much value at all.
This example stands out: Eat a balanced diet focused on fresh fruits and veggies, whole grains and healthy (non-saturated) fats. https://www.cdc.gov/healthyweight/healthy_eating/index.html