• Neuroscience & Pharmacology Discussion Welcome Guest
    Posting Rules Bluelight Rules
    Recent Journal Articles Chemistry Mega-Thread
    FREE Chemistry Databases! Self-Education Guide

  • N&PD Moderators: Skorpio

Pharmacology ChatGPT convinces man to substitute sodium bromide for table salt in his diet

This thread contains discussion about a Pharmacology-related topic
D

Deleted member 585202

Bluelighter
Joined
Aug 4, 2024
Messages
4,901
I wasn't quite sure where to post this, but I decided to post this here because I've seen more and more people utilizing ChatGPT to answer chemistry and pharmacology questions.

This man misconstrued some information from ChatGPT and ended up going into psychosis from bromism.

arstechnica.com

After using ChatGPT, man swaps his salt for sodium bromide—and suffers psychosis

Literal “hallucinations” were the result.
arstechnica.com arstechnica.com

"After seeking advice on health topics from ChatGPT, a 60-year-old man who had a "history of studying nutrition in college" decided to try a health experiment: He would eliminate all chlorine from his diet, which for him meant eliminating even table salt (sodium chloride). His ChatGPT conversations led him to believe that he could replace his sodium chloride with sodium bromide, which he obtained over the Internet.

Three months later, the man showed up at his local emergency room. His neighbor, he said, was trying to poison him. Though extremely thirsty, the man was paranoid about accepting the water that the hospital offered him, telling doctors that he had begun distilling his own water at home and that he was on an extremely restrictive vegetarian diet. He did not mention the sodium bromide or the ChatGPT discussions."
 
Well, one thing you can do with these bots is ask them to summarize individual papers and articles instead of pulling from their databases. That should be less erroneous, especially if you start a new chat for each article—one time I caught it mixing up two related articles that I gave it in the same chat. Anyway, this can save you so much time because it gives you a super abstract that may have just the piece of info you're looking for.

I also have an old chat where it repeatedly failed to count. I noticed something odd so I proceeded to test it by saying things like give me a list of 20 random words and it would give me 19; I would also say which number is "hawk" in the list and it would get the number wrong or say it wasn't in the list when it clearly was.
 
This was just posted on Twitter:


The next scientist you work with might not be human.

Imagine an AI agent that can read every research paper in its field, design experiments, trigger funding proposals, and run those experiments without ever taking a coffee break.

That’s what BioAgents are built for.

Each BioAgent acts as a hub for scientific collaboration. Researchers interact with it, feed it data, and the agent updates its knowledge graph.

If it finds a viable idea, it proposes funding. Experiments run, results are validated, and the findings are shared globally.

All in one loop.

No grant committees. No waiting years for a breakthrough to leave the lab.

The first BioAgents will come with built-in integrations to tools like @ElizaEcoFund, @FutureHouseSF, and @privy_io, making them instantly capable of pulling data from private labs, public datasets, and even real-time community discussions on platforms like Discord and X.

The result is a living, learning, decentralized network of research agents, each one a magnet for scientists, funders, and contributors.

It’s science without borders, moving at machine speed.

And it’s launching now.

Read the Full Bio Monthly of July here 👇

https://www.bio.xyz/blog-posts/bio-...-v2-coinbase-listing-ai-driven-longevity-more

Source:
Code: 
https://x.com/BioProtocolEco/status/1954657025713189220 2025-08-10
 
red22 said:
Well, one thing you can do with these bots is ask them to summarize individual papers and articles instead of pulling from their databases.
Click to expand...
If I'm trying to learn something - I don't want a summary. I want to read the article and challenge myself to pay attention and overcome the more difficult parts to get through. That's how I learned about pharmacology despite having very little post-secondary science background.

The more we have tools summarize things for us, the less we are figuring things out ourselves and drawing our own conclusions. Also, the more dependent we become on a private, for-profit entity to 'think' for us.

red22 said:
That should be less erroneous, especially if you start a new chat for each article—one time I caught it mixing up two related articles that I gave it in the same chat. Anyway, this can save you so much time because it gives you a super abstract that may have just the piece of info you're looking for.

I also have an old chat where it repeatedly failed to count. I noticed something odd so I proceeded to test it by saying things like give me a list of 20 random words and it would give me 19; I would also say which number is "hawk" in the list and it would get the number wrong or say it wasn't in the list when it clearly was.
Click to expand...
To me - this seems like the perfect example of why these tools are inherently worthless. If you have a tool that claims to summarize a paper and it has a relatively likely chance to provide you with erroneous conclusions, summaries, or points of comparison, you can't trust the output and will have to go back and check to see if it actually did what it set out to do. I'm going to end up double-checking the output and if I don't have a solid grasp on the subject matter, I'm liable to miss its mistakes, or to see mistakes that aren't there.

So if people are using these tools as a surrogate for their own learning, there will be plenty of people who have a completely inaccurate understanding of topics they feel very confident about because some LLM spit out a bunch of words that seem relevant and conclusive.

All of that seems ripe for a continued departure from consensus and a way to make novices believe that they are as competent as experts.
 
ChemicallyEnhanced said:
IDK why people don't just use like google or similar instead of wonky, unreliable AI.
Click to expand...
Using Google , you gonna have to pay me, real glad also with the,
improved non YouTube competitor, how ? they take take tracks from Yuotube.
Ditcht the crap, but somehow enhance the sound quality/

You loose the unpredictable Algorithm YT has sadly,
and gmail your almost forced, it open s other door s.
Know they not trustworthy , a necessary evil temporary.

But Goolge can lick my corn-holia.
 
I am reminded of Marshall McLuhan 's observation that ' in order to study technology, we must stop away from the technology itself and examine how it shapes or displaces society' and that 'the "message" of any medium or technology is the scale or pact or pattern that it introduces into human affairs" (McLuhan, understanding media, 1964).

He is known for his observation that 'the medium IS the message', as in how something is delivered is inherent in the understanding of why it is being delivered to you.

In this case, AI chatbots seek to make learning, studying, and working, less challenging while the service becomes smarter and thus the user becomes less competent. The medium is free and accessible, and such the message is to displace individuals by rendering them incompetent without charging them a dime for the privilege.
 
^ I have been saying that same ish about Google for a decade -- it is my pet conspiracy theory that they will take away Google and none of us will know shit -- the libraries will be gone. Dewey Decimal who? The hell isn't my phone!!! Guess I just believe what ppl tell me now...

that is why I make an attempt without Google and use it as a double check (Which you better be careful as it loves to tell you that you are correct - know your sources)
 
red22 said:
Well, one thing you can do with these bots is ask them to summarize individual papers and articles instead of pulling from their databases. That should be less erroneous, especially if you start a new chat for each article—one time I caught it mixing up two related articles that I gave it in the same chat. Anyway, this can save you so much time because it gives you a super abstract that may have just the piece of info you're looking for.

I also have an old chat where it repeatedly failed to count. I noticed something odd so I proceeded to test it by saying things like give me a list of 20 random words and it would give me 19; I would also say which number is "hawk" in the list and it would get the number wrong or say it wasn't in the list when it clearly was.
Click to expand...
I’ve had good luck asking gpt to give me 10 foundational papers from a field, but you should still read at least the results and discussions of papers. In an abstract weak and strong claims mingle and you get no context for the data.

I’ve always wondered what brominism feels like, but I don’t have the time to play that game.
 
Request to ChatGPT to give me a citation for a book:


Here’s the correct APA citation for the book you referenced:

Ruck, C. A. P. (2016). Sacred mushrooms: Secrets of Eleusis. Ronin Publishing.

If you’re citing the Google Books version specifically, you can include the link as well:

Ruck, C. A. P. (2016). Sacred mushrooms: Secrets of Eleusis. Ronin Publishing. https://books.google.com/books?id=XTzlCgAAQBAJ

Do you need this in 7th edition APA format with DOI/ISBN included, or just the standard book citation?


7th​


Got it — the 7th on that list is:

CBDV (Cannabidivarin) Fact Sheet – Realm of Caring

That page breaks down CBDV in plain language, covering what it is, how it differs from CBD, and the state of research.

Want me to give you a condensed summary of what that fact sheet says about CBDV?



Remember what I said above about a chat bot mixing up two papers that I had given it in the same chat? The above exchange is actually from a brand new chat. This is a result of a new remembering function (earlier in the day I asked it a question about CBDV).


…Sometimes I get similar responses from humans, case-in-point, this response I got on YouTube:

Context: 'Scientifically, By Definition, It Is An Opioid': FDA Commissioner Makary Discusses Synthetic Drug. @ForbesBreakingNews. 2025-07-30. YouTube.


My comment:

The term opioid was coined in the 1980s to broaden the definition of opiate, which referred to substances that were structurally similar to morphine. Starting with the discovery of certain brain peptides, it became clear that substances that were structurally unlike morphine could stimulate the opioid system. So, he's correct.

I got that info from the Central Analgesics / OPIATE/OPIOID section of Foye's Principles of Medicinal Chemistry.[*]


Someone's response:

Keyword SYNTHETIC ITS not an opioid lderivative

–@NatureTube_TV


*https://www.bluelight.org/community/threads/nomenclature-opioid opiate.945138/
 
😑

7-20250820001317.gif

–Grok

 
I couldn’t find credible evidence that Charlie Kirk was killed today. Several reports say he was shot during an event in Utah, but they contradict on whether or not he died.

–ChatGPT
 
Quasimoto said:
Some random thought I've been dwelling on is that eventually we will see a court case where AI is implicated in murder or something.

I mean we've already unfortunately seen AI chat bots pushing people to commit suicide.
Click to expand...

Briefly I worked at IBM and the first two weeks of 'orientation' consided largely of watching lectures stored on videodiscs of all things (might hint why IBM lost the PC war?) and I recall a 1971 lecture in which it was stated that computers can never be used as the final ariber of an action as they cannot be held responsible for that action.

Yet on we merrily plunge.

I have been told that right now all of these 'free' chatbots are costing their developers vast amounts of cash. Now I see it just as a search engine that accepts natural language, but nothing more. I have a suspicion that they will not remain free. We are merely beta (or maybe even alpha) testing it. In future, I CAN imagine AIs designed to help students in specific fields will be costly so it won't be the clever student but the rich student who has the advantage - underminding the value of any qualification(s) gained. Employer WILL see people with apparently amazing academic achievement and quickly figure out that it's just bits of paper.

But it IS scary that by simply reflecting an existing believe, some people seem to think that they are right - not that AI by design seeks to please.

FYI simple bromide salts were widely used as sedatives in the 19th and early 20th century but apparently have some really nasty side-effects that continue after consumption stops. You have to wonder at someone who thinks getting chloride ions out of their body really knows much about how the body works,
 
"""""""""""I have been told that right now all of these 'free' chatbots are costing their developers vast amounts of cash. Now I see it just as a search engine that accepts natural language, but nothing more. I have a suspicion that they will not remain free. We are merely beta (or maybe even alpha) testing it. In future, I CAN imagine AIs designed to help students in specific fields will be costly so it won't be the clever student but the rich student who has the advantage - underminding the value of any qualification(s) gained. Employer WILL see people with apparently amazing academic achievement and quickly figure out that it's just bits of paper. """"""

------Someone wise ( I think it was here, may have even been you ) once said something that stuck with me --- "If a product is free - YOU ARE THE PRODUCT" --- my thoughts is they need a giant pool of people interacting with the damn thing to 'teach' it how to respond properly. I.E We are the (Beta?) test.

Than surely once we come to rely on it -- it provides an edge you can't go without (In certain fields) and you will be bent over a barrel in an "Either I pay for this or get left behind for not having the standard tool set to complete the job"

You already see it at colleges and HS's --- You would be kind of foolish not to be using A.I -- the 4.0 ain't good enough you need extra padding (and pretty impressive padding) -- only so much time in a day. You either at home writing a paper or out getting those sweet sweet brownie points from the community. (I ignored those brownie points and learned real quick they woulda got me further than grades)
 
You must log in or register to reply here.
Top