GROK just exposed itself - Built by BlackRock
(twitter.com)
Comments (16)
sorted by:
Ive caught Grok lying on several occasions. When i say lying i mean fabricating narrative- supporting “facts” out of thin air, when Grok knows those facts are false.
Every AI does that.
Yesterday I simply asked to do a dive on the Green word: Priton - Pritones.
AI posited the the 5th century Greeks were accomplished navigators, thus calling the people living on the, what we now would call: British Isles: Britons, or Pritones, which refers to their cultural characteristic of meeting and discussion before making a decision.
So, I requested the sources where AI got his ideas from. There are no sources. Its a constructed narrative based on the idea of being aware of a people means contact, meaning: Herodotus was in contact with these people. Which is ludicrous on first sight. Herodotus even misnamed: Egypt, Lybia, the Nile, etc.
Then it became clear that the prevailing narrative of the books of Homer being a Greek story was the basis for its assertions. Which is ludicrous on its face.
I countered it by requesting an explanation how Menelaus could end up in Egypt, thus south of Rhodus, while being driven by a southern storm. I asked if AI new about other sailors of them days having the same experience?
And here you see the idiocy of AI.
It is fun to use, but is misses something, something that cannot be copied: Human experience! Why? Because human experience is beyond parameters and objectives.
The best AI so far, is my own trained AI. And actually, if you really want a good AI, you need to build it from the ground up:
Do you let your ai build go online at all? If so, how do you control what online sources it uses?
No online connection. I determine what it can and cannot read. This way, I prevent contamination.
Basically, the salient is point is not the vast body of bullshit created by bots ... I feed it the salient points of humanity.
I can then contrast it with several versions of AI, like perplexity, and see what comes from it.
In the end, I am the decider, and not AI.
One of the issues with AI is, that a word or syllable is just a representation of a number. Based on probability and statistics it tries to predict the next set of letters. That the outcome makes sense to us, is of no consequence. It simply performs according to an algo.
For instance: the word: de-finit-ion basically means: moving away from the borderline. So, when we inquire after what a matter is, we want to use such word as to put it in a border, a frame, otherwise, this does not differ from that.
So, in essence, we are using a word with a meaning of the **opposite ** of what we want and do. In regular life, we do not give a rats ass about it, of course. However, there is reason to be circumspect, because of the nature of the universe on a quantum level expressed with the words: be careful what you wish for.
It is just one example of cognitive dissonance in daily life.
AI has a hard time to comprehend that as a word or syllable represents calculations, a mathematical expression that needs following. It is void of meaning in and of itself. We assign meaning to it!
Mathematics is called: the language of God. I am not sorry to say: bollocks.
Mathematics is a model, an approximation that may or may not hold true. I can see the attraction of modelling.
Instead, I am trying to get my locally running AI to take in salient point to which to measure certain aspects. The issue I am running into is the use of certain words of Latin origin that are meant to be mis-used due to the combination of prefixes and suffixes giving the word a vector away from the thrust it should and ought to have.
Very interesting
Oh you have. I just thought Grok was not very thorough and missing info.
I've said this before but I believe that AI will follow the same path as Internet search engines. They will start off quite simply but will gradually improve over time. People will come to rely on them and then the results will start being "manicured".
The answers they want you to see will appear and the ones you don't will be buried. The AI program will sound certain about some things and less sure about others. Everything out of an AI programs needs to be checked.
When all you can do is check the output against a search engine we have a problem. Wasn't it supposed to have been the CIA that said that when everything the people believe is not true then we will know that we have done our job?
It's true. I used it yesterday night and I found out some really sad case especially in communist countries. You don't even get the information you want; not just on page 27, but not at all.
Nonetheless, I tried Grok and didn’t like at all. ChatGPT agrees, naturally:
Interesting. Thank you for this.
When 3 corporations own everything, it’s unavoidable
Exactly. This isn't even news.
When 3 families control Rothschild and Soros, it's inevitable.
Black Grok
Yes, definitely BlackGrok.