Signal already has that setting. It’s up to the user to decide their level of convenience vs security.

data security in that case had nothing to do with the llm

That’s kinda my point.

“I don’t trust companies to hold their promises” is a very different argument from:

LLMs are inherently bad at data security and there is no way these companies can, in good faith, promise HIPPA compliance

It is certainly possible to implement a secure LLM service.

Phone, wallet, keys, and headphones

This is about extracting data that was used as training data. Just don’t do that with sensitive data.

LLMs are inherently bad at data security and there is no way these companies can, in good faith, promise HIPPA compliance

This is simply false. AI sucks but it doesn’t help to lie about it.

EDIT:

Go run a local model on your own computer, and delete the context when you are done. Boom you just used an LLM in a way that maintains your data security.

I have a bare private git repo on my homeland server. Not great for sharing my work but great for personal projects.

LLMs tend to be a “jack of all trades, master of none”. You are likely to find them useful for helping you with something you are inexperienced at, but not at something you are an expert in. However, because they lie a lot, it’s best to double-check your information, but the LLM can still be helpful with the ”you don’t know what you don’t know” issue.

Two weeks is about how long I took when I made that trip, but I did some sightseeing on the way. Also even on the longest day I don’t think I drove for 10 hours in one day.

Natural disaster -> no longer can access everything you have online, including bank and insurance accounts, at precisely the time you most don’t want to deal with that.

Somebody do this to Minecraft

More specifically I thought one of the approaches to an omni-treadmill would catch on enough for an at-home model to be available to the public.

Previous admin was pro-Ukraine but this admin is pro-Russia

It’s the same price and similar specs to current Chromebook models, which is what I think they are trying to compete against.

Yes, there are difficulties in the design of experiments and studies sometimes. Things like control groups and placebos are designed to rule out certain very common confounding variables. If you cannot have a placebo, you might still be able to get useful data by other means. For example, sometimes comparison to an existing drug is used instead of comparison to a placebo.

Ultimately it all comes down to statistics. Typically, you start with “assuming” the “null hypothesis” (basically that you are wrong). For example: that your medicine doesn’t work and/or has bad side effects. Your goal is to find evidence to reject that null hypothesis with sufficient confidence. This can be done by any means, but statistics should be your guide, and you have to be careful about bias and confounding factors, and standard study formats and advice are tried-and-true reliable methods to avoid common issues. But if those don’t work for some reason, it is ok to get creative, as long as your math checks out.

If you can’t run a standard study, you should try coming up with a creative study. If you can’t come up with a way to correct all the issues, you might try studying related topics. If you really can’t gather meaningful information about your topic, that’s tough but I absolutely reject the idea that you should take something as true without true evidence just because it’s too difficult to get that evidence.

In your specific example of corpus callostomy, I would bet that 100% of cases where this surgery was performed were well documented, including follow up visits. That’s fantastic for your statistics, and means you don’t have to worry about a lot of sampling issues that you would otherwise have to correct for. You might not be able to perform experiments or new studies on the topic, but you can certainly learn from the documented cases, and you can look at studies on related topics like brain injuries, or experiment with animals (the ethics of that is a whole other debate).

An example of how this kind of reasoning works (note that I’m making up the specifics here): 100% of people who got this surgery had a post-surgery event where left-and-right hands fought. It seems like this is related to the surgery, but we have to be sure it’s caused by the surgery and not just some confounding factor like the symptoms that cause people to get this surgery in the first place. So we do a study of people who have symptoms that would have qualified them for the surgery, but instead get a different treatment or no treatment. If none or very few of those people have left/right arm fights, then we can say we have sufficient evidence that this symptom is caused by the surgery.

This is very different from the NDE topic, in which a huge number of people suffer near-death situations, and only a tiny fraction of those end up with supernatural experiences. We want to prove these supernatural experiences are real, but the incidence rate is so low it could just be statistical noise. To show evidence of the supernatural you’d need some way to demonstrate that it’s not just statistical noise or other “mundane” / “null hypothesis” explanations.

I want to mention a more science-y topic that fits into this pattern I read about the other day. If you are interested let me know and I’ll try to dig up the sources.

There is a significant amount of neurons throughout the body (outside the brain). One particularly large collection of those is in the heart. This is sometimes called the “brain of the heart” and is in charge of controlling the heart muscles with only high-level instructions from the brain. There was a hypothesis that some other behavior might happen in that heart-brain such as storing memories. This idea came from a couple case studies where a heart transplant recipient would seem to gain memories or personality traits from the donor. These cases sounded a lot like the typical “paranormal knowledge” story. Two particular cases were someone liking a food they didn’t like before but the donor did, and a child avoiding a toy that donor had with them when they died. Personality change is common after transplants in general, presumably because of the immense stress and changing life habits related to the situation. So a study was done, where they interviewed a selection of transplant recipients of both the heart and other organs and recorded any personality changes to see personality changes in general, or if some specific types of personality changes, were more common among heart transplant recipients than others. The results showed that the only statistical difference between the heart and other organs was personality changes related to sports or exercise, which has the much more mundane explanation of being a result of the symptoms of having an y healthy vs healthy heart.

Disproving ideas is just as important as proving them. That’s the whole reason for the scientific process: to make sure what we accept as fact is very likely to be fact.

Not exactly “misheard”, but the lyrics to the Minecraft parody of “Dynamite” stuck more strongly in my head than the original.

You can’t turn pure heat into useful energy. Thermoelectric generators tap into the transfer of heat between a hot reservoir and a cold reservoir.

Literally everything you mentioned has at least had its ban discussed, and most of those have been banned or at least restricted in some part of the world.

I think a 1% chance of permanent health effects manifesting years later is already plenty to get something banned.

DDG works 90% of the time but it does perform worse than Google sometimes