AI chat bots programmed to provide misleading medical and scientific advice

Has anyone found that AI chat bots provide intentionally incomplete, incomplete or outright false information on a variety of topics, including medical, scientific and political?



I have been researching how much truth is behind the generally accepted dietary recommendations, which apparently propagate a mindset of "unless you get this much of this and that much of that you are in danger". This is in spite of the fact that inuit tribes beyond the arctic circle have been living for hundreds of years without ever consuming ANY fruit or vegetable as nothing grows in the eternal ice and snow there. Surely you cannot get your "5 a day" from just meat. Same goes for Pacific island tribes (that only have fish and coconuts), people living high in the mountains in China living on rice, or African tribes living in the Kalahari and Namib deserts. I am talking about pre-industrial revolution times with severely limited transport and commerce, when trading was minimal or non-existent in these areas.

So I decided to try the most readily available AI chat bot that requires no registering and see what information it will provide about a(n apparently not-so-)well-known Guinness record holder who was published in a number or prestigious medical journals for not eating anything for more than a year under constant medical observation.

What I got was lie after lie after lie:



This would've gone unnoticed by 99.9% of the public, except I have copies of the original journal publications to compare the responses with:



Note that he went for more than a year with ZERO vitamin A intake. Wasn't he supposed to go blind, or suffer from skin problems?

I have had several similar conversations only to find a staggering amount of censorship, obviously misleading the public on matters of anything from health to politics.

Has anyone else found this? Is this being addressed? I fear these AI bots will be presented to the masses as trustworthy arbiters of truth, when they are anything but. The perfect brainwashing machinery, making school obsolete soon, or even news channels and independent media. In fact I see how this might tie into the prophesied dissolution of all religions. Am I alone with these views?

a reply to: Rolci

I share your doubts about recommended daily allowances and food groups. Plenty of mistakes have been made, and since we're not perfect, will continue to be made. Just look at all the ridiculousness regarding fats, and especially dairy fats. Generally speaking, I think a varied diet including basically equal amounts from all food groups is optimal, but there will always be exceptions.

AI's limitations and inaccuracy in this regard is expected: garbage in, garbage out.

But regarding the clinical reports on the guy fasting, he was receiving Vitamin A and/or beta carotene via the Multiviteand Paladac supplements. I would expect that his vitamin/mineral levels were closely monitored, and at the point his electrolytes began to suffer, they began supplementation with potassium and then sodium.

The problem is that you, and most people, can’t comprehend that language modeling does not equate to comprehension. There is nothing intelligent about AI. AI is a misnomer and gullibility is the layman’s ignorance.

originally posted by: Rolci
Has anyone found that AI chat bots provide intentionally incomplete, incomplete or outright false information on a variety of topics, including medical, scientific and political?



I have been researching how much truth is behind the generally accepted dietary recommendations, which apparently propagate a mindset of "unless you get this much of this and that much of that you are in danger". This is in spite of the fact that inuit tribes beyond the arctic circle have been living for hundreds of years without ever consuming ANY fruit or vegetable as nothing grows in the eternal ice and snow there. Surely you cannot get your "5 a day" from just meat. Same goes for Pacific island tribes (that only have fish and coconuts), people living high in the mountains in China living on rice, or African tribes living in the Kalahari and Namib deserts. I am talking about pre-industrial revolution times with severely limited transport and commerce, when trading was minimal or non-existent in these areas.

So I decided to try the most readily available AI chat bot that requires no registering and see what information it will provide about a(n apparently not-so-)well-known Guinness record holder who was published in a number or prestigious medical journals for not eating anything for more than a year under constant medical observation.

What I got was lie after lie after lie:



This would've gone unnoticed by 99.9% of the public, except I have copies of the original journal publications to compare the responses with:



Note that he went for more than a year with ZERO vitamin A intake. Wasn't he supposed to go blind, or suffer from skin problems?

I have had several similar conversations only to find a staggering amount of censorship, obviously misleading the public on matters of anything from health to politics.

Has anyone else found this? Is this being addressed? I fear these AI bots will be presented to the masses as trustworthy arbiters of truth, when they are anything but. The perfect brainwashing machinery, making school obsolete soon, or even news channels and independent media. In fact I see how this might tie into the prophesied dissolution of all religions. Am I alone with these views?

Which AI program is this one? I suppose it's free to use. I want to ask it what does it think about the woke culture.

This is a true test for artificial intelligence..

It's best to read and compare sources. Font trust the AI. Even if it's right on some matters.

"Has anyone found that AI chat bots provide intentionally incomplete, incomplete or outright false information on a variety of topics, including medical, scientific and political?"

No and I suspect that anyone else who thinks LLMs are 'intentionally' doing anything would change their tune if they just spent even a little bit of time reading about what LLMs are, how they work, and what are some of the challenges going forward (hallucinations - or what some awkward people call "lying", for example).

I don’t think he is implying the language model is actually thinking on its own and lying/omitting.

I believe he means the people who made the model made it to do those things. The language models omit by intent. The intent of the creator.

a reply to: ASrubWhoDiedAgain

Peoples' bodies have adapted to the food available in their region through the centuries. As an example, I live on Nez Perce reservation land and have a front row seat to what a modern diet has done to their health.

Their original lifestyle had them put on weight in the summer months when food was plentiful, and through the winter months they would shed excess weight. With a modern diet the 'feast and fast' cycle is broken and they are now plagued with once unknown diseases to them, diabetes and renal disease being epidemic. Their bodies are also far more susceptible to the ravages of 'fire water' and regular measles can be life-threatening to Native Americans.

My friend, who holds a degree in biology, explained that it takes approximately 400 years for a group to adapt to a foreign pathogen.

Some people don't digest meat easily, of which I'm one. Alpha-gal syndrome (a tick-borne disease) is beginning to sweep the US; those affected become allergic to meat, often with life-threatening reactions 3-8 hours after consumption.

When you add in all the chemicals used to grow, process and preserve our food all the FDA health recommendations don't mean squat to me. If they aren't going to take the whole picture into account in their analysis I won't be putting any stock into their ever-changing dietary requirement updates.

Average life expectancy pre-industrial revolution was like 39 years old.

originally posted by: Athetos
I don’t think he is implying the language model is actually thinking on its own and lying/omitting.

I believe he means the people who made the model made it to do those things. The language models omit by intent. The intent of the creator.

a reply to: ASrubWhoDiedAgain

That belief too stems from ignorance towards how the tool actually works.

There are weights for sure (built in biases, the technology literally can't work without them) and there is to a degree some model chaining (parse output from model 1 for the N word and return a pre set enum if present) but to suggest that across a wide range of topics responses are 'intentionally' omitted requires yet another model to deduce and this becomes a recursive issue that requires an infinite amount of compute or a self training model that can do so without weights (impossible to achieve atm and have function as intended).

Models are trained on content produced by people. People are generally bad at stuff.

a reply to: ASrubWhoDiedAgain

There are tons of YouTube clips that explain how they work. I found the graphics provided in most made it alot easier to understand.

Problem is, we struggle to understand their own coding.

AI chat bots programmed to provide misleading medical and scientific advice

AI chat bots program provides ridiculous medical and scientific advice.
It's the programming that is flawed and is not deliberate.
It's just the way it works , as it is not AI .

a reply to: Rolci

Depends what data the AI 'learned' from: as always, "Garbage In, Garbage Out."

originally posted by: Rolci
Has anyone found that AI chat bots provide intentionally incomplete, incomplete or outright false information on a variety of topics, including medical, scientific and political?

If many are trained off the Internet and esp. social media, this would be no surprise at all.

What I have found is these bots talk too much. Like the smartest person in the room who needs to make certain you know their status. They can't just give a clear, concise answer. They also don't take well to being told they're wrong.

originally posted by: Boadicea
I think a varied diet including basically equal amounts from all food groups is optimal, but there will always be exceptions.

Pretty varied here too. I get my grains and greens pre-processed through though those things that mooooooo.

originally posted by: Deplorable

Pretty varied here too. I get my grains and greens pre-processed through though those things that mooooooo.

I had to smile when I read this, but that's a perfectly cromulent means of getting your nutrients! Especially if it's well-marbled... that fat contains lots of nutrients.

Those magical animals that make ham/bacon/sausage is also a good source. Conventional wisdom may frown on that lovely creamy delicious substance called "lard," but besides flavor, it adds a whole lot of nutritious goodness to food.

originally posted by: Terpene
a reply to: ASrubWhoDiedAgain

There are tons of YouTube clips that explain how they work. I found the graphics provided in most made it alot easier to understand.

Problem is, we struggle to understand their own coding.

Try to use them here when you promote your ideologies. Someone else tried but it didn't work well because the programme was written probably by a like-minded person. It exposes both the woke and the woke AI

originally posted by: quietspectator
Average life expectancy pre-industrial revolution was like 39 years old.

You do know that's because the rate of sudden infant death syndrome (SIDS) was high in those days, bringing the average down, and not because most adults lived to around that age, right?

a reply to: Rolci

Why on earth would anyone go to an AI bot for medical advice?

It's not discerning, it can't make judgments about whether information is good or bad or from a reliable source. It's about as accurate as any random googling.

This theory isn't so far out there if you go back to the very beginning of the internet.

The internet began in 1966 as a US Department of Defense project called ARPANET. According to the former director of DARPA, it "was to exploit new computer technologies to meet the needs of military command and control against nuclear threats, achieve survivable control of US nuclear forces, and improve military tactical and management decision making." The very internet itself is a military tool for command and control. Now think about where we are today with the internet and the integration of AI into this system, think about the far greater command and control capabilities of the government.

AIs aren't programmed, but they do make mistakes. Like here. Also, who be going to chat bots for anything other than generative hallucination? Certainly not for important data. So all we be left with here then is future doom porn.