Forum: Author Hangout

Illegal since colonial times.

I didn't read that in that article. In fact, the book referenced was the 3rd reference in the Meta AI response that I didn't bother to include.

Meta AI said that although it was legal, there were many groups against it and prostitutes were arrested for crimes like vagrancy and disorderly conduct. Not specifically prostitution. Prostitution in NY State wasn't illegal until 1921 (according to Meta AI).

From an article that is actually talking about the economics and land value in NYC due to prostitution: https://digitalcommons.iwu.edu/cgi/viewcontent.cgi?article=1027&context=uer#:~:text=Although%20violence%20against%20brothels%20was,'disorderly%20persons'%20.%20.%20.

Although violence against brothels was prevalent in the early 1800s, regulation of prostitution proved unsuccessful before 1870. Prostitution itself was not considered illegal, however the "police could and did arrest 'all common prostitutes who have no lawful employment' as vagrants or 'disorderly persons' . . . (in addition) though the police periodically raided Sixth Ward brothels and often hauled in streetwalkers from predominately immigrant areas, elite brothels were almost never disturbed."18

Footnote 18 is: "18 Burrows, Edwin G., and Mike Wallace. Gotham. (Oxford: Oxford University Press, 1999) 807."

But my post wasn't about prostitution in 1870s NYC. It was that I got conflicting answers from 2 AI engines.

Both their ability to synthesize results and their tendency to hallucinate make them unreliable for that type of research.

Yeah, that's what I surmised. There's so much incorrect information on the internet that it makes sense that since that's the source for the AI programs "learning," the results should be questionable.

If you trust Wikipedia at all, you are making a leap of faith

This was mainly caused by how stupidly easy it was to 'edit' articles when the site first went 'live'. It was a pranksters wet dream and boy, did they not gleefully accept the challenge.

Since then, they (the site owners) have made it harder to edit articles and edits now need to be approved/validated before being added to the main site.

These days, the site is every bit as reliable as DTP encyclopedias. More so, as mistakes in print will only be corrected with the subsequent edition, whilst mistakes in Wikipedia are corrected as soon as they are noticed.

It's not yet 100% trustworthy, but then, neither is the MSM, which is always issuing retractions and apologies because of mistakes with their reporting. Compared with MSM, Wiki is more trustworthy, and as for those amongst us who read technical/medical/research papers, downright lies are frequently to be found, written by those who should really know better.

This is why teachers and professors always say to both quote the source AND double check the data.

These days, the site is every bit as reliable as DTP encyclopedias

No.

A year or two back, a newspaper article exposed how easy it is to get bad info onto wikipedia. You write a 'paper' in support of your opinion, you get it published in a 'vanity' scientific journal (of which there's been a huge proliferation, particularly on-line only), you cite your paper in your contribution to Wikipedia (spit!) and, hey presto, your info has become gospel.

AJ

These days, the site is every bit as reliable as DTP encyclopedias.

Maybe for things that are nearly impossible to politicize.

Wikipedia is now a leftist propaganda machine. Anything that has any political potential is heavily skewed left. Historical items have been rewritten to serve the current agenda.

Even Jimmy Wales, Wikipedia's founder, admitted that Wikipedia has lost its credibility because of its extreme leftist bias.

The tricky part is that it's a billed as a reference source. Who needs a reference source? People who don't know the info and need it. If you trust wikipedia, then you'll be misled on countless topics. The only people who know what can be trusted or not are the people who don't need wikipedia on the topic.

has lost its credibility because of its extreme leftist bias.

Ditto for almost any news media on the internet (and TV) so you can't seem to get the truth anywhere. So you might as well use wikipedia. For that matter, it's true for what's being taught in our education institutions, especially at the university level. So what is the truth?

Even Jimmy Wales, Wikipedia's founder, admitted that Wikipedia has lost its credibility because of its extreme leftist bias.

I looked for this quote, but couldn't find it. What I did find is:
"Jimmy Wales, a co-founder of Wikipedia, stresses that encyclopedias of any type are not usually appropriate as primary sources, and should not be relied upon as being authoritative." from a 2005 article in Businessweek. [This info should be obvious advice to anyone doing research outside of well-establish 'first' sources]

Please support your assertion.

I looked for this quote, but couldn't find it.

It was in a recent interview that he made.

I don't recall where I've seen the video, it was probably on Twitter/X.

Please support your assertion.

I didn't find a quote by Jimmy Wales, but I found this article "Wikipedia co-founder says site is now 'propaganda' for left-leaning 'establishment'" quoting Wikipedia co-founder Larry Sanger.
https://nypost.com/2021/07/16/wikipedia-co-founder-says-site-is-now-propaganda-for-left-leaning-establishment/

Now keep in mind, "The NY Post" is a right-leaning newspaper.

Wikipedia co-founder Larry Sanger has warned that the website can no longer be trusted — insisting it is now just "propaganda" for the left-leaning "establishment."

If you trust Wikipedia at all

AIs use social media as part of their input. Social media is even worse than Wikipedia's input.

I don't disagree. That sounds like the major problem with an AI's output.

Personally, I think the main problem goes back to the software programmers who created the AI program. Then there are the people who accept the program as generating accurate viable output.

My granddaughter is doing a Master's in a medical field. Many of her fellow students use an AI program to write the reports that they have been assigned as an aid in their learning. Just think about the quality of medical treatment their patients will receive. You and other Forum participants may one day be one of their patients.

I can just picture them inputting the patient's symptoms and asking an AI program for a diagnosis.

I can just picture them inputting the patient's symptoms and asking an AI program for a diagnosis.

I think that's where the future of General Practice lies. There's too much information for one person to remember. Some conditions can go undiagnosed for years because individual GPs are a long way from knowing everything. Is there a better solution than a trained practitioner using a purpose-built AI to access an on-line system containing all the latest information?

AJ

I agree. In my opinion, a medical text would be the preferred source of a diagnostic guide.

In the case of my granddaughter's friends, they are likely to not know enough about their specialty to use a medical text, and if they can, they are likely to not understand what they read.

Good God Almighty!

I just checked and the medical field is introducing AI into making diagnosis. It is believed to be the future in the medical field. However, remarks are being made that AIs are not ready yet and when they are, human intervention is necessary. Currently, AIs are to take on the role as an assistant to the doctor.

I doubt I would trust a doctor who uses an AI to diagnose my ailment.

I doubt I would trust a doctor who uses an AI to diagnose my ailment.

AIs are particularly useful when they are doing something boring, like analysing X-rays or cervical smear samples, because unlike people, they don't get bored and their attention doesn't wander.

AJ

If they can't keep the AI from inventing a fictional diagnosis

That's why I said purpose-built. You obviously wouldn't use an existing AI and train it from the internet.

It's already starting to creep in in the UK. When the doctor is entering the symptoms and diagnosis, they get asked whether they're sure it isn't sepsis or cancer. Allegedly most doctors ignore the prompts.

AJ

No, purpose-built does not let you hand wave away the problem of AI fabricating information from nothing.

LLMs fabricate information from nothing. Special-purpose neural nets generally don't.

The problem here is people talking about 'AI' is if it's one thing. It's not. It's many different, highly divergent, types of software.

A purpose-built LLM will fabricate in some (but not all) cases. There are some which only use the LLM on the input side, using it to select between a large number of human-written outputs. Those don't fabricate. They may give you the wrong output if they misunderstand the question, but the output will be correct (and is hopefully written in such a way as to make clear what it's answering).

The problem here is people talking about 'AI' is if it's one thing. It's not. It's many different, highly divergent, types of software.

Some claim that telephone answering systems, that keep you going round in circles rather than letting you talk to a human when your problem isn't covered by any of its options, are AIs. (Not mentioning British Telecom by name)

AJ

I have yet to encounter a VRU (Voice Response Unit) that wouldn't transfer me to a live person in the situation where my issue is outside it's programming.

I have.

I have yet to encounter a VRU (Voice Response Unit) that wouldn't transfer me to a live person in the situation where my issue is outside it's programming.

I have too. An insurance provider who REALLY wanted people to use its website. It was supposed to attach you to a real person at the second time of choosing 'car insurance' but that didn't work.

AJ

Sadly, I'll second the others. I've met at least three of them. One of them let you leave a number for a live person to call you back at some arbitrary point in the future, but I left a number, got no callback, googled it, and found out that nearly no one every gets a callback.

They're far more common in 'support' situations than 'sales' situations, for obvious reasons.

The solution to getting a live person at two of those companies was to call sales and make them transfer the call. Surprisingly or not, that worked.

Tried 1 and 2 (those are among my go-to's).

3 is good, when you can. Sometimes you can't, or it's so impractical that it's worth putting up with the crazy.

I can just picture them inputting the patient's symptoms and asking an AI program for a diagnosis.

This is inevitable, but they won't be chat / LLM AIs, they'll be neural nets trained on disease symptoms.

There are already AIs who can take certain easy-to-obtain screenings and detect things no human could detect with more than ample accuracy to justify followup screenings, and we're still (relatively speaking) in the infancy of such tools.

Neural nets are great at finding very subtle patterns in data. The problem, of course, is that one has to be careful about what is the 'signal' and what is the 'noise'.

There's a great story about a program that was shown a large number of pictures. Half of them had tanks (the armored vehicle sort) in them; half did not. The program became very, very good at identifying tanks.

Then they showed it a new set of tanks and it failed miserably. After some investigation, it turned out that the 'with tank' photos were taken in one season, the 'without tank' photos were taken in another, and what they had was an AI that picked between the two seasons.

Still, get your inputs right and you'll wind up with AIs that are better than nearly any human doctor at making diagnoses from a set of symptoms.

Anyone trying to do that with an LLM is a fool, though. That's not what LLMs are for.

They may be, one day, but that'll take another feedback look that can semantically understand the answer and fact-check it. We're not there yet, but the path to getting there is much clearer than it would have been not that long ago.

The great thing about Wikipedia is it has links.

It's also it's weakness because it doesn't vet the quality of the underlying documents. With traditional encyclopaedias, humans supposedly check the quality of the supporting information.

On the other hand, Wikipedia(spit!) is now far beyond typical dead-tree encyclopaedias, both in the range of articles and their being up-to-date.

AJ

But you can vet the quality of the source yourself.

In real life, how many people routinely do that? I think most people assume that because there is a link, it somehow verifies whatever is claimed.

AJ

That could well happen if medical AIs were let loose on teh interwebs using Large Language Models. But medical AIs are more like Lazeez's wizards.

AJ