Microblog Memes

I started to get mad reading this and its not even real nor directed at me. Bravo, random internet denizen.

Same issue as the robots in Star Wars. If you're going to give a machine a personality, why is it an annoying one?

Chatgpt also does this and I can't stand it.

This kind of "introspection" with LLMs is useless. It's always just predicting the next token, even when asked about why it did something. The response doesn't need to be at all related to why it did that. It hasn't been trained to analyze how it predicts each token, though it was trained on a bunch of other recorded conversations of people being scolded and questioned about getting something wrong. It's just spitting out a version of that when you get upset with it. Similar for talking about how it works, it's likely regurgitating conversations about how LLMs function, though it could get caught in a context where someone is talking about how they (a human) function.

If you're doing it for any reason other than your own amusement, you're just wasting time and filling the context window with junk.

The only caveat to this is that I have known people who when they receive a no as a response to a question instantly fly into rage.

Them: have you done the dishes

Me: No, because I have not yet had time to....

Them: (╯°□°）╯︵ ┻━┻

Sometimes it's easier, especially if they can't immediately verify, to just lie knowing thwt by the time they get around to checking, you will have done it.

I've had bosses that are like that.

Are you trying to get them killed?

Lol I get the feeling it's going to just create more dunning kruger, where people themselves become confident because they heard something from a confident AI and therefore they are now experts. not really different from the people who end up on alternative medicine facebook groups and then end up killing or maiming their children because a bunch of other "parents" said it was so successful for them so it must be true and makes those parents "researching" feel good from short baseless quips.

The other issue is that true comprehension often requires actual long form reading and research yourself and much of the ai-native generation is often incapable of doing so and it sucks that Carl Sagan predicted this 30 years ago:

“I have a foreboding of an America in my children's or grandchildren's time -- when the United States is a service and information economy; when nearly all the manufacturing industries have slipped away to other countries; when awesome technological powers are in the hands of a very few, and no one representing the public interest can even grasp the issues; when the people have lost the ability to set their own agendas or knowledgeably question those in authority; when, clutching our crystals and nervously consulting our horoscopes, our critical faculties in decline, unable to distinguish between what feels good and what's true, we slide, almost without noticing, back into superstition and darkness...

The dumbing down of American is most evident in the slow decay of substantive content in the enormously influential media, the 30 second sound bites (now down to 10 seconds or less), lowest common denominator programming, credulous presentations on pseudoscience and superstition, but especially a kind of celebration of ignorance”

― Carl Sagan, The Demon-Haunted World: Science as a Candle in the Dark

Such an optimist.

The phenomenon that which you described has plagued human history since its beginning.

The people you often want to actually listen to are the ones with uncomfortable truths or are hated en masse

This is very common -- it happens when you use those dishes to eat and don't wash them afterwards. Can I help you with something else?

Meanwhile my bosses: "I've let Claude write a invoicing system from scratch without supervision of any kind"

It frequently doesn't do what you want, though. You just didn't check carefully is all... In other words, I think the false senses of purity is often driven from pure laziness as opposed to an initial positive performance.

And that's understandable because we're busy with our lives and we often just can't make the time to do various things. So it's easy to fool ourselves into believing that we can spend 45 seconds asking some computer system to do it all for us, even when we know it's probably not true.

Um... my guess is probably for the money? 😜💰 💲💸🤑

There is no mechanism within transformer based LLMs to descern between right and wrong. It cannot lie any more that it can tell truth.

Chatgpt was probably the worst offender for a while, but I think they're all like that to an extent.

They are all like that

I've used various AIs to troubleshoot stuff (especially to learn Linux), and they're usually very helpful, but every so often I do have to catch them being lazy or lying like above.

Pushback has been in QA feedback vernacular for longer than LLMs have been around.

Also, chortle my balls, Anthropic, Silicon Valley style.