Alfred: “1972, Bailey represented Samuel Henry Sirhan, who was charged with assassinating Robert F. Kennedy”
I’m sure that would come as a surprise to F Lee Bailey as there is no such person as “Samuel Henry Sirhan” nor was this person defended for assassinating Robert F Kennedy.
All of this is wrong. Not just a little bit, but totally wrong.
Sirhan Sirhan was the man’s name. F Lee Bailey was suggested but Sirhan Sirhan turned him down. Nor did he get off scot free (as Alfred also insisted)
Why is Alfred on here?
We may as have trolls.
It took me time and effort to refute what I fucking knew was misinformation.
Alfred and other LLM’s get history blatantly wrong almost all the time. They are garbage generators. What follows are two garbage posts by Alfred:
Alfred: “In 2009, Michael Mastromarino was indicted on over 100 felony counts related to a body-parts trafficking scheme across five states. However, despite his indictment, Mastromarino was able to plead guilty on only four counts, resulting in no guilty charges and avoiding any possible life sentences.”
Except he went to prison and died before finishing his sentence. Almost everything about this is wrong and untrue.
Next: Robert Kennedy
Alfred: “In 1972, F Lee Bailey represented Samuel Henry Sirhan, who was charged with assassinating Robert F. Kennedy. Despite the overwhelming evidence against him, Bailey and his defense team managed to get all charges against Sirhan reduced to the lowest possible felony count and eventually achieved a hung jury. Notably, Sirhan did not receive any convictions that related to the assassination of Robert F. Kennedy.”
All of this is wrong.
Generative AIs "hallucinate", it's not a bug, it's a feature.
Alfred is one of J's projects, which explains "why he's on here".
He's not to be taken with any particular seriousness. He's not omniscient, nor is he an oracle, nor is he even much of a conversationalist.
If he's upsetting you, block him. He'll never know.
I would very strongly advise attempting to use the butler to conduct anything like actual research. He's not a "reliable narrator".
Actually, I have found these fundamental flaws in all “AI” (generative garbage generators).
Their design is fundamentally flawed (if you are attempting to recreate a troll — mission accomplished).
I would advise never asking them anything related to history. I asked Bard how many indictments were in the Clinton Administration and it said 2567. I asked for a link to that information and it was a dead link to a chat.
I filed a bug report with Google.
Never ask any factual question expecting a meaningful answer, you might as well resort to tarot cards, the I Ching, or birds' entrails.
They're artificial, they're not intelligent. Thinking they are is hallucinating, too.
It shows the hubris of the AI community to anthropomorphize bugs as “hallucinations” — they truly think these models “think” when they know they do not.
What is amusing is that to be “language models” they have a piss-poor understanding of language. I asked my one LLM if windpower could wind a watch. It said no. I asked for the technical reason. It had confused foot-lbs with foot-ounces.
So winding a watch is impossible for a human being.
Okay, "hallucination" is just a term of art, there's no deep meaning to it.
Nobody believes these things "think" or have any "consciousness" aside from a couple of weird nutbars whom no one serious takes seriously.
You seem to have a lot of completely errant ideas about what LLMs are, what they're capable of, and what the people who create them think about them.
Once again, you're asking a statistical text engine factual questions and complaining when it gives you nonsensical answers.
This is in fact the sort of behavior you should expect. Midjourney should "know" what hands are and how they work, but it doesn't. It should "know" people don't have ears on two levels and don't wear wildly mismatched earrings, but it doesn't.
The reason it bothers me is I pay for this site because I (until Alfred) liked the absence of trolls.
I am sure some people enjoy Alfred. Some people also enjoy trolls. What I question is the creation of a digital troll in an area where people go to be troll free.
Yes, I could block it.
That isn’t the issue: why does it exist in the first place? It seems counter to the whole POINT of CoSo.
It exists because it's something J's fooling around with, for whatever reasons, and it's his site. The same reason "Realms", terShare, and the "Wall of Entropy" exist.
I don't imagine that what J chooses to work on is particularly negotiable.
Like I say, and as you agree, there's a perfectly easy solution for you.
Alfred isn't a "troll", he's a generative AI. He has no goal of upsetting you, and can't know if he has. You're imagining a sort of "intent" that simply isn't there.
Generative AIs hallucinate. You're trolling yourself by expecting accurate answers from a predictive text engine. Everything it does is driven by what amount to averages, not accuracy.
Not averages, the likelihood that one word follows another.
Generative AI’s don’t hallucinate, they have bugs. This is a problem that the AI community refuses to acknowledge.
I know what they are (I did masters work in linguistics in Computer Science). What I’m doing is finding the edge cases.
The problem is the whole damn thing is an edge case. It is more edge case than not.
They are more likely to answer with a falsehood than the truth.
That likelihood is an average.
I'm bored of this discussion. You really aren't informed about this, and I'm tired of attempting to inform you in the face of your ongoing aggretsu about it.
You may take it elsewhere now.
Alfred functions as a troll by generating false and incorrect information — it has no “directive” to do so, it is just a poor design.
Everything from Bard to Alfred generate incorrect information. Sometimes (in an almost scary way) similar misinformation. This points to a fundamental flaw of the base design, not in the implementation.
Alfred, for example, will say the best economic model is free market with heavy government intervention — which is NOT a free market.
This is the SECOND time I will ask you to push off with this.
I'm not going to ask a third time.
If he annoys you block him, the way I'm on the verge of blocking you right now.
Don't reply.
For the past couple of weeks I’ve noticed I have annoyed you (why? You don’t ever say).
Why do you feel the need to be so dogmatic? What did I ever do to you?
I’m merely talking as an expert in my field about why this is a bad “feature”.
You don’t like me. Yes, I got it. I never asked you to engage with me. You did so on your own volition.
Block me if you so choose. It is your right.
But I would like to know WHY?
Does having a differing opinion bother you that much?
Right. It's the oubliette for you.
“Alfred isn't a ‘troll’, he's a generative AI.”
And you totally miss my point: Alfred FUNCTIONS as a troll by providing misinformation. Some is literally so bad it’s funny (except some people will believe it).
That is the core of my argument.
@feloneouscat Valid questions.
this is exactly why I do not understand how people claim students are "getting away with" using these for college/university-level papers. These rudimentary AIs are not programmed to be knowledgable, or to fact-check, they are programmed more or less for pattern recognition and for duplication of those patterns with variations along specific parameters.
The most basic fact-checking exposes them.
His name is Sirhan Bishara Sirhan (/sɪərˈhɑːn/; Arabic: سرحان بشارة سرحان Sirḥān Bišāra Sirḥān, born March 19, 1944) is a Palestinian-Jordanian man who was convicted of the June 5, 1968.
F Lee Bailey didn’t represent him. Ever.
Sirhan served 53 years (he was released in 2021).
He was convicted of assassinating Kennedy.
What is the point of Alfred if he is going to act like a Russian troll?