"I just found out that it's been hallucinating numbers this entire time."
-
@Natasha_Jay So the jobs of the VP and CFO were useless to begin with? I'm not surprised.
-
@Natasha_Jay just a reminder that an IAgen doesn’t tell the truth. It tells a plausible answer. That’s what « generative » stands for.
-
@Natasha_Jay Hallucinating is still the wrong word, it's a needless anthropomorphism, they are simply invented numbers...
-
@Natasha_Jay This reminds me of this post - sorry no link - where some talked about introducing CoPilot into his company, for 1000 people, when only 7 people use it actively and he boss can now say that they are AI-enabled company!
-
@Natasha_Jay
I am sure that, as usual, this will be used as an argument against AI, not as one against Natural_I. -
@Natasha_Jay note the complete absence of gasps following that statement
-
Obviously, whoever authorised the implementation should be sacked for failure of due diligence.
In a previous life though, I remember talking to someone about their data. "Is this accurate?" I asked. "It's plausible" was the response. A discussion followed around a viewpoint that it isn't actually cheating if you know everyone else is cheating in the same way. Which is why it makes sense not to believe that the figures issued by any organisation are completely accurate, they'll often be presented in a way that supports a particular narrative.
Which isn't what you were talking about, but it did remind me, sorry!
@Fragarach @Natasha_Jay "it isn't actually cheating if you know everyone else is cheating in the same way"
Which is why anyone who might eventually be prosecuted as a result of being a customer of the Epstein child rape gang is going to feel singled out and hard done by.
-
@Natasha_Jay No one could have predicted this.
-
@Natasha_Jay I love this particular use of AI - disrupting capitalism with a plausibility bomb.
AI was really developed by the Anarchist Illuminati to bring on the Nonsensemageddon.
-
What would be amusing is them having greater success using bullshit data than whoever was previously correlating stuff :^D
I can't find it now, online search has become almost useless.
But there was a news story some years ago where they compared predictions from 10 stock market tipsters. Then they added a mystery 11th to the mix.
When the results came in,the mystery 11th tipster came in a close second if I remember right. The mystery 11th turned out to be a pet parrot. -
@Natasha_Jay "I just found out it's been bullshitting the whole time."
-
@Natasha_Jay It's like a maths professor who played the lottery for years, and then one day sat down to actually look at it...?
-
@Natasha_Jay AI failing upwards just like any good manager
-
I can't find it now, online search has become almost useless.
But there was a news story some years ago where they compared predictions from 10 stock market tipsters. Then they added a mystery 11th to the mix.
When the results came in,the mystery 11th tipster came in a close second if I remember right. The mystery 11th turned out to be a pet parrot.@Fragarach @lxskllr @Natasha_Jay Maybe this one? Monkeys not parrots. https://www.marketsentiment.co/p/the-monkeys-that-beat-the-market
-
@ktneely @lxskllr @GreatBigTable
I think AI is technically a scrape goat.
-
@Fragarach @lxskllr @Natasha_Jay Maybe this one? Monkeys not parrots. https://www.marketsentiment.co/p/the-monkeys-that-beat-the-market
@robinadams @lxskllr @Natasha_Jay
That's really interesting, thank you for that!
The one I'd been thinking of was much less rigorous and thorough. I had no idea so much effort had gone into those comparisons. -
@Natasha_Jay Or they made bad decisions and are now blaming an LLM, because this excuse can help dodge responsibility. It's like blaming an intern.
-
@Natasha_Jay An organization which trusts this tech enough to simply implement without a changeover period during which the results now given are getting checked against old methods... simply deserves what it gets.
@ZenHeathen @Natasha_Jay but the new numbers are so much better!!!! 🤪
-
K don't understand what has happened here. Please help me.
One of my jobs was crunching data for senior management reports. A job that would be automated today. Producing metrics implies you are transforming raw numbers into meaningful trends and forecasts.
Where was the invention? Was it in the input or transformation? Both should have guardrails. In other words the ai could, for instsnce, be using the wrong statistical analysis to produce the wrong results. But that's an error not an hallucination. People do it too.
I've had hallucinations but if you require ai to show sources you soon spot made-up stuff. Trouble is when the sources are unchecked ai generated (ai slop). That's just beginning.
