Coconut Monkey Cornerclub

[Zed Clampet]

@BeardyHat @Pifanjr zedllucination correction: I think I was specifically talking to you to when I had my zedllucination. I said that I thought the reason for LLM hallucinations was that there was a conflict between what they were designed to do and how they are being used. I said that they were designed to run simulations.

Well, actually: Earlier AI had been designed for simulations, but, according to CoPilot, the purpose of the LLM's was to create an Artificial General Intelligence. Even as of only a year or so ago, the AI folks believed that if they just kept feeding LLM's training data that they would eventually be AGIs. All hope for that has been lost.

So why do hallucinations occur? Don't know. All I have is the traditional answer, which everyone knows, which is that they don't know anything. They are just selecting the next language token based on the internal logic they developed while training. I have problems with this explanation, but the people who espouse this actually know what they are talking about, so I'll keep my theory to myself rather than switching on Auto-Idiot (TM) and typing away as usual.

 

[Zed Clampet]

@Frindis @Colif and other people interested in the absolute best blocker:

Google despises Ublock Origin and completely redid parts of its code in order to get rid of it. There are "lite" versions available for Chrome, but it isn't nearly as robust as it used to be.

Please note that there are fake versions of Ublock out there. If websites are still asking you to turn off your ad block, then either you don't have the real Ublock or you have a version that's been dumbed down to work with the new Chrome.

The browsers that fully support Ulbock are Firefox and Brave. Edge and Opera still fully support it, but the assumption is that they will stop support soon.

 

[Colif]

Firefox is the obvious choice as all Chromium based browsers will eventually break them. Firefox also lets you pick which AI you want to use... I just ignored the choice so I don't have any helpers... oh no, what will I do?

I haven't tried Brave so cannot comment. Future of Firefox is unclear, hope they keep surviving as I would hate to have to stop using internet completely.

Chrome should be taken off Google and made into its own company completely apart from the Advertising business. Then they might not hate Ublock as much... but now, they sell advertising and obviously hate adblockers for that very reason. How dare people use the free internet and not see their ads... or let us track peoples habits.

 

[BeardyHat]

@BeardyHat @Pifanjr zedllucination correction: I think I was specifically talking to you to when I had my zedllucination. I said that I thought the reason for LLM hallucinations was that there was a conflict between what they were designed to do and how they are being used. I said that they were designed to run simulations.

Well, actually: Earlier AI had been designed for simulations, but, according to CoPilot, the purpose of the LLM's was to create an Artificial General Intelligence. Even as of only a year or so ago, the AI folks believed that if they just kept feeding LLM's training data that they would eventually be AGIs. All hope for that has been lost.

So why do hallucinations occur? Don't know. All I have is the traditional answer, which everyone knows, which is that they don't know anything. They are just selecting the next language token based on the internal logic they developed while training. I have problems with this explanation, but the people who espouse this actually know what they are talking about, so I'll keep my theory to myself rather than switching on Auto-Idiot (TM) and typing away as usual.

I assume hope has been lost because now they have billions of parameters and are pretty much the same as they've been for a couple of years now?

Apropos to this, this weekend, we went on a mini road trip, 3 hours south into New Mexico to meet a sister in law who lives there. We met at a park, which my wife surprised me with because it had a monument tank. So after lunch my kids and I went and climbed all over it and I took lots of reference pictures for my future paint projects.

Anyway, I wasn't sure what it was. It had a hull very similar to a late war US tank (M41), but was equipped with what appeared to be an anti air gun. So I popped it into Copilot on my phone and asked if it knew and it told me it was an M4 Sherman, which was just so egregiously wrong, I ended-up just searching Google instead and found my answer that way.

Turns out it was an M42 "Duster"

M42 Duster - Wikipedia

en.wikipedia.org

 

[BeardyHat]

Actually, I should point out that the cool thing we did, aside from climb on a tank, was hike an ancient volcano. Down into the caldera and up along the rim; it was super interesting to see the shape of the cooled lava flows from the top, which are still evident these millions of years later.

 

[Zed Clampet]

I assume hope has been lost because now they have billions of parameters and are pretty much the same as they've been for a couple of years now?

Is there some metaphysical moment where suddenly it's read so much 4Chan that it just springs to life like Frosty? I almost think that's what some of them think. But, you are probably right. The change in attitude happened when they fed the AI the last "lol" from the Internet and nothing happened.

Someone is going to figure out something better than LLMs. It may even be a combination of electronics and living brain cells. Now that is what will kill us all 🤣