EpeeGnome

When a headline asks a question, then the answer is usually “It’s not that simple, but, no, not really”. Sounds like it applies here. “Well, they often were painted, but no, we have no reason to assume they did a bad job like we sometimes do.”

Edit: The post title was changed to a statement, but the article is still titled “Were classical statues painted horribly?”.

When I was in college, in a network programming class we had a semester final coding assignment. I forget now what all it was supposed to do, but I do recall my friend in the same class spent a week of free time writing his. I forgot about the assignment completely until he asked me how it was going for me a few hours before it was due. With no other options, I simply punched in a example program from the textbook, replaced about 10 lines in the middle and tested.

It seemed to work. So, I very clearly labeled which code was mine, and which was copied, so that I technically wasn’t doing a plagiarism. Then I turned it in, and hoped for the best. I got a 100 on it. My friend was pissed, because he only got a 99.

Huh, never tried it, and never done it by accident that I noticed. Makes sense though. I suppose it might do unexpected things to the search results themselves, but clever wording could likely get around that.

AI doesn’t have the ability to critically think

That is absolutely correct, and the deeper problem is it typically “writes” in a tone that fools people into feeling like it can do so.

just outputs what it can find on the internet

That is what this one, the Google search overview, is set up to do. Other LLMs don’t, or may only do so when prompted to.

What they all do is analyze the patterns of all the words that have already been input, such as the initial system prompt, the user’s prompt, any sources it’s referencing and any replies it’s already generated, and then it uses that to predict what words ought to come next. It uses an enormous web of patterns pulled from all of its initial training data to make that guess. Patterns like a question is usually followed by an answer on the same topic, a sentence’s subject is usually followed by a predicate, writing tone usually doesn’t change, etc. All the rules of grammar it follows and all the facts it “knows” are just patterns of meaningless symbols to it. Essentially, no analysis, logic, or comprehension of any kind is part of its process.

This is the Google search overview AI. It just reads the top results and summarizes them together. You don’t directly prompt it, it’s already prompted to just do that. The problem with that arrangement, as demonstrated here, is it will confidently and non-critically summarize parody, idiotic rambling, intentional misinformation and any other sort of nonsense that the search algorithm pulls up.

One of the reasons I switched to a PieFed server. I’ve been happy with the switch though, it feels generally smoother in my browser as well. That’s the beauty of the Fediverse. Even on a whole different platform, I’m still able to participate in this discussion hosted on a Lemmy server.

“AI should always be a choice—something people can easily turn off." “It will evolve into a modern AI browser and support a portfolio of new and trusted software additions.”

How does he not get how contradictory these positions sound. Really a missed opportunity to brand themselves as the browser without AI bullshit and gain users who want to get away from that crap. Sure, they promise it’ll have an off switch, but even if that’s true, they’re still wasting a lot of their very limited budget pursuing it. Really shows where their priorities are.

My high school only had one pay phone. It had a bad connection in the hand set, so sound cut in and out constantly. People rarely ever bothered making calls on it. The coin return also had some sort of obstruction inside it. If you inserted a quarter and then hit the coin return lever, you’d hear it fall, but it didn’t actually come out. When enough quarters built up though, they would all flood out into return tray at once. Naturally, it got used as a slot machine. Drop in a quarter, pull the tiny lever, and see if you hit the jackpot.

If all 6 got the same answer multiple times, then that means that your query very strongly correlated with that reply in the training data used by all of them. Does that mean it’s therefore correct? Well, no. It could mean that there were a bunch of incorrect examples of your query they used to come up with that answer. It could mean that the examples it’s working from seem to follow a pattern that your problem fits into, but the correct answer doesn’t actually fit that seemingly obvious pattern. And yes, there’s a decent chance it could actually be correct. The problem is that the only way to eliminate those other still also likely possibilities is to actually do the problem, at which point asking the LLM accomplished nothing.

This is why all capital 'i’s should have bars on the top and bottom. It is my strong opinion that disambiguating serephs should not be considered optional, even in sansereph fonts.

Yes, I just glossed over that detail by saying “similar to”, but that is a more accurate explanation.

“They’ve been saying so many bad things about him for decades, and nothing stuck. He says he’s innocent, so who should I believe, the man I trust, or the lying media and the deep state? They just have it out for him because he tells it like he sees it.” You can point out all the inconsistencies in what Trump says, but they’ve tuned out already. They already know what they need to about him, or they just aren’t detail people at all. Some are so far gone, they think, “He’s a good person, therefore whatever it was he did was a good thing.”

Unfortunately the most probable response to a question is an authoritative answer, so that’s what usually comes out of them. They don’t actually know what they do or don’t know. If they happen to describe themselves accurately, it’s only because a similar description was in the training data, or they where specifically instructed to answer that way.

I particularly enjoyed Adobe® Creative Commons™ software and Microsoft Minecraft® Bedrock™ for Workgroups™.

Wow, this takes me back. This was one of my favorite sets to play with, and I had forgotten it. I was always a little disappointed that the magnet arm couldn’t actually reach the little magnet cargo box in the base. I usually put the robot on the torso-docking hover-platform, and the human in the head/spaceship, on account of the human being safer in an enclosed cockpit, while the robot was less at risk riding exposed to the vacuum of space. I also added some upgraded thrusters to the head from some other set so it could fly faster when detached.

Thanks for reminding me, I’ve been meaning to do this. I’m doing random text edit of them all now. I’ll go back later and have it edit my top comments to a nice site-ban-worthy message.

Yeah, this really depends on what you meant by winning here. Are we actually changing the other person’s mind or do they admit we are right, but still believe otherwise? Are people who witness the argument included? Do people continue to agree indefinitely? Does it change reality to match?