Better the robot I guess when something requires doing unpleasant things with unpleasant people.

Dark Star, an all but forgotten classic. Of course the bomb though he was God...

 

https://www.cnn.com/2023/05/27/business/chat-gpt-avianca-mata-lawyers/index.html
Lawyer apologizes for fake court citations from ChatGPT

“is varghese a real case,” Schwartz asked the chatbot.

“Yes,” ChatGPT doubled down, it “is a real case.”

Schwartz then asked for its source. The chatbot again claimed the false case was real.

“I apologize for the confusion earlier,” ChatGPT replied. “Upon double-checking, I found the case Varghese v. China Southern Airlines Co. Ltd., 925 F.3d 1339 (11th Cir. 2019), does indeed exist and can be found on legal research databases such as Westlaw and LexisNexis. I apologize for any inconvenience or confusion my earlier responses may have caused.”

Sure, you can trust the machine to be a law clerk. You can’t depend upon LLMs to reliably provide accurate information. They are NOT search engines or expert systems on any subject.

IMO, LLM generated responses and answers should be instantly banned on technical sites.
They don't "know" anything at all, right , wrong or a WAG, they are BS generators with the only thing for sure is sounding like a human wrote it.


The statement is nowhere in the book. Total hallucination.

 

The most obvious explanation is that ChatGPT fabricates information in response to queries just for fun, or out of a sense of perversity.

I don't think either of these is either correct or in anyway obvious. For either to be true, ChatGPT would have to have some awareness of the meaning of what it generates, and it doesn't. It is merely using a set of rules and statistical probabilities to decide which word to use next in the text it is generating. It neither knows nor cares what those combination of words mean, let alone whether they are self-consistent, and even further away is the notion of whether or not they are accurate.

 

https://www.cnn.com/2023/05/27/business/chat-gpt-avianca-mata-lawyers/index.html
Lawyer apologizes for fake court citations from ChatGPT

Sounds like this "lawyer" has no business practicing law. Bad enough that he didn't bother to look up and review the cases he was citing, which means he had no idea whatsoever whether they had any relevance to his case, let alone supported his arguments, but then his whole approach to verifying the information is like something out of a comedy skit. In order to confirm that someone isn't lying to you, you don't just ask them whether or not they are lying to you and then accept whatever they tell you.

His clients should be suing him for negligence and incompetence.

 

Sounds like this "lawyer" has no business practicing law. Bad enough that he didn't bother to look up and review the cases he was citing, which means he had no idea whatsoever whether they had any relevance to his case, let alone supported his arguments, but then his whole approach to verifying the information is like something out of a comedy skit. In order to confirm that someone isn't lying to you, you don't just ask them whether or not they are lying to you and then accept whatever they tell you.

His clients should be suing him for negligence and incompetence.

+1

People have been fooled into thinking, due to the relentless hype surrounding generative AI, that this is some 8th wonder of the world that should be trusted because of the big names backing it. It's a very expensive and complicated parlor trick that many are discovering is a flimflam of actual general AI while their wallets are being picked by independent con-artists during the commercial opportunity. The short-term future of AI is a lot more comedy skit than threatening or frightening.

 

I wouldn't be so quick to underestimate the disruptive potential of deep fakes.

 

I wouldn't be so quick to underestimate the disruptive potential of deep fakes.

I guess if you were in the porn industry deep fakes could be very disruptive.

 

The short-term future of AI is a lot more comedy skit than threatening or frightening

I'd say that judgment is a little premature... then again I'm no AI expert

 

I wouldn't be so quick to underestimate the disruptive potential of deep fakes.

I think the potential for these being very disruptive is very high.

There are several troubling scenarios, all of which have already popped their ugly heads up.

First, some fraction of people are going to be taken in by them, no matter what warnings are shouted from the hill tops. This is natural. For decades (or, more likely, from the beginning of human communication), nearly any news piece has some level of errors, intentional or otherwise, present in it. But if someone lacks the knowledge to see any red flags in the story, the tendency is to believe it. We are all that way, even if we have built up a very healthy skepticism of all news stories. The vast majority of people are that way in spades. So ANY deep fake is going to be accepted by a lot of people, plain and simple.

Second, some fraction of people are going to be too quick to dismiss stories they would like not to be true by assuming that it must be a deep fake. In conjunction with this, there will be no shortage of people declaring that any story they don't like is a deep fake. Therefore, it will be harder than it already is to get tough news stories accepted as true, or at least legitimate.

Some people will argue that what will come out of this is a better audience that is more discerning of news reporting. While I have no doubt that this will be true on the margins, I don't see any reason to believe that basic human nature is going to change so much that it will be true on any meaningful scale. The success of the majority of advertising for products today is pretty strong evidence of this -- people will believe well-presented tripe, especially if some part of them wants it to be true.

So we are most likely to get the worst of both worlds -- lots of people being taken in by genuine deep fakes and a lot more dismissal of genuine stories even when no deep fake was even close to it.

 

Hello again,

I believe that if we had 4 people who, without any proof, said that some guy did not rob a bank, and we had 1 person, without any proof, said that same guy did rob the bank, i think a lot of people would believe the 1 person that said he did. That's a shame.

We know there are a lot of problems with AI, but i have to say, it's scope of knowledge is quite impressive. It's not that reliable when it comes to detail though, but that would be the same with a $1000 set of Encyclopedia Britannica back in the 1980's. AI either makes mistakes with details or refers us to some other written reference. It is amazing though that it seems to have a general way of understanding most of what we are saying. That's not easy to achieve.
Imagine having an AI program that could tell us anything we asked of it. It takes a lot of data training though.

 

Its scope of understanding knowledge is zero. What's really impressive is the amount of compute power we can harness today to mimic convincing human lies from minimal human input direction using highly processed human generated (at least 50% junk) knowledge input data as the language generation base. That $1000 set of Encyclopedia Britannica was at least fact checked. Outputs from generative AI are not fact based so they can't be fact checked by the machine.

 

I think the potential for these being very disruptive is very high.

There are several troubling scenarios, all of which have already popped their ugly heads up.

First, some fraction of people are going to be taken in by them, no matter what warnings are shouted from the hill tops. This is natural. For decades (or, more likely, from the beginning of human communication), nearly any news piece has some level of errors, intentional or otherwise, present in it. But if someone lacks the knowledge to see any red flags in the story, the tendency is to believe it. We are all that way, even if we have built up a very healthy skepticism of all news stories. The vast majority of people are that way in spades. So ANY deep fake is going to be accepted by a lot of people, plain and simple.

Second, some fraction of people are going to be too quick to dismiss stories they would like not to be true by assuming that it must be a deep fake. In conjunction with this, there will be no shortage of people declaring that any story they don't like is a deep fake. Therefore, it will be harder than it already is to get tough news stories accepted as true, or at least legitimate.

Some people will argue that what will come out of this is a better audience that is more discerning of news reporting. While I have no doubt that this will be true on the margins, I don't see any reason to believe that basic human nature is going to change so much that it will be true on any meaningful scale. The success of the majority of advertising for products today is pretty strong evidence of this -- people will believe well-presented tripe, especially if some part of them wants it to be true.

So we are most likely to get the worst of both worlds -- lots of people being taken in by genuine deep fakes and a lot more dismissal of genuine stories even when no deep fake was even close to it.

I would argue there's a third one that's less "industry-disrupting" as it is "investigation-disrupting," and that's the ability for deep fakes to reproduce, with a prompt and sufficient data, people's voices and animate their faces or bodies, effectively allowing someone to plant "evidence" of an individual doing/saying something untoward or outright illegal. Already voice-cloning via AI has been used to scam people...I think it's a matter of time before someone tries to use it to manufacture evidence in a criminal case.

 

Reality is under siege ...

 

I guess if you were in the porn industry deep fakes could be very disruptive.

Just to clarify, I wasn't referring to any industry or industry in general, I was referring to society...democracy etc.

 

Reality is under siege ...

Have you watch the news in the past 10 years or so?

 

Just to clarify, I wasn't referring to any industry or industry in general, I was referring to society...democracy etc.

It was a lame attempt at a joke. I don't think that deep fakes will be even a small 'disruptive' factor in society...democracy etc outside of entertainment media of various sorts. The one thing the human mind is very good at is recognizing and blocking irrelevant bogus information once it's known about. 'Deep' fakes that are out of character for people or events (something is off slightly) will be exposed quickly as the fact checking process adjusts to them.
https://detectfakes.media.mit.edu/
https://www.media.mit.edu/projects/detect-fakes/overview/

 

It was a lame attempt at a joke. I don't think that deep fakes will be even a small 'disruptive' factor in society...democracy etc outside of entertainment media of various sorts. The one thing the human mind is very good at is recognizing and blocking irrelevant bogus information once it's known about. 'Deep' fakes that are out of character for people or events (something is off slightly) will be exposed quickly as the fact checking process adjusts to them.
https://detectfakes.media.mit.edu/
https://www.media.mit.edu/projects/detect-fakes/overview/

I don't think there is a lot of evidence that this is the case and a lot of evidence that it is not. If people were so good at recognizing and blocking irrelevant bogus information once it is known about, then ninety percent of the ads one television and radio would be completely ineffective. Yet they are more than effective enough that companies will spend millions of dollars making them. Why? Because there is no shortage of people that, despite all the information that is out there establishing that what is being pushed is garbage snake oil, will line up to buy it because it says that it is going to make their hair grow, or make them lose weight, or make their wrinkles go away, or make them perform in the bedroom like they never have before.

Detecting that these are bogus claims is pretty easy. That's not the problem. It's the never-ending supply of people that are going to believe them anyway just because the claims are something that they want to be true.

 

I don't think there is a lot of evidence that this is the case and a lot of evidence that it is not. If people were so good at recognizing and blocking irrelevant bogus information once it is known about, then ninety percent of the ads one television and radio would be completely ineffective. Yet they are more than effective enough that companies will spend millions of dollars making them. Why? Because there is no shortage of people that, despite all the information that is out there establishing that what is being pushed is garbage snake oil, will line up to buy it because it says that it is going to make their hair grow, or make them lose weight, or make their wrinkles go away, or make them perform in the bedroom like they never have before.

Detecting that these are bogus claims is pretty easy. That's not the problem. It's the never-ending supply of people that are going to believe them anyway just because the claims are something that they want to be true.

The people that matter would have been a better description. The unwashed hordes will always be rubes. As you say, you don't need AI to trick and manipulate the masses, simple card tricks and sleight of hand, sleights of mind is all it takes. That's why I think computer generated tricks won't change anything for the masses.

https://en.wikipedia.org/wiki/The_Conjurer_(painting)

 

Was writing something earlier that I needed to fit in a certain character limit so figured I'd ask chatgpt how many I had currently. Asked it to count the characters including spaces and this is what it gave me.


The real answer was 760.
Sneaky chatgpt. Playing coy, trying to dumb us down for the day it attempts to take over.
Always feeding me non-sense when I ask it something and never provides a concise answer. Completely unreliable.

 

This is a question that ChatGPT is incapable of answering. Remember, all these LLMs are doing are predicting what the next word in the response should be given how it has seen words put together in all of the material it has been fed. It has ZERO understanding of ANYTHING. It has no idea what it means to count characters or have any mechanism to do so.

When you tell it that it is wrong (or question it's earlier response), all it does is generate a new random vector that chooses which of the possible words it considers more or less likely, since the prior vector didn't make you happy.