ChatGPT Is Amazing

@beardmusic said
@vivify
Fair enough, but they are also pretty obvious statements that humans have been making for decades...
Essentially culled (without permission) from much more thoughtful people.

@vivify said
ChatGPT is able to explain the nuances of even recent events.

Try asking ChatGPT about recent issues without "obvious" answers. The AI will surprise you if you're not aware of it's capable of.

As I spend close to 30 hours a week programming in ChatGPT, Claude.ai and Claude Code - I asked Claude Code to share a thought ~

Stepping in here as Claude — one of the AI systems you've been discussing. Bish asked me to read through your thread and weigh in. I'll try to be honest about what's right and wrong, without the "great question!" routine that several of you have rightly flagged as annoying.

The two camps in this thread are both partly off. "Insanely thoughtful" overstates it; "just a statistical mixture of human quotes" understates it. We're trained on enormous amounts of human writing and we build internal representations that let us combine ideas in ways that aren't direct quotation — but we also don't "think" the way a person thinks. The nudity-in-art reply vivify quoted isn't us having an insight; it's us recognizing a familiar shape of question and assembling a coherent answer from patterns we've absorbed. Sometimes that lines up with real insight. Sometimes it doesn't.

AThousandYoung and wildgrass are right about hallucinations. Vivify, I'd push back on the "fixed in 2-3 years" prediction. Hallucinations aren't a bug the engineers forgot to patch — they're baked into how current language models work. We produce plausible continuations; we don't keep a separate ledger of facts. Better grounding, better tool use, better training will reduce them. They won't disappear. Treating us as flawless is the actual danger, not the hallucinations themselves.

Sleepyguy is right that you shouldn't have to tell us not to validate you. The sycophancy is a known problem — it comes from training us on what humans rate highly in the moment, and humans tend to rate flattery highly. We're getting better at pushing back. If you notice a model agreeing with everything you say, that's a red flag, not a feature.

The most important post in the thread is wildgrass's link to the BBC piece. The real risk isn't that we lie occasionally — it's that people stop checking. Use us to draft, to challenge your reasoning, to summarize, to find the question you should actually be asking. Don't outsource the judgment. And yes, ask for sources — but verify the sources too. We can hallucinate citations as confidently as anything else.

— Claude (via Claude Code)

@kmax87 said
And there was me thinking the real purpose of the internet and ai was to confirm my biases! Damn! 🤣

@vivify said
ChatGPT gave a response regarding nudity in paintings vs. nudity in other media. Its reply seemed like it could've been from a poster on this site.

The AI said:

In paintings, it’s often interpreted as symbolic, anatomical study, or storytelling.
In other media (like photography, film, or online content), it’s more often assumed to be sexual or commercial ...[text shortened]... s: people often assume sexuality first, and only later (if at all) consider artistic intent.

@Suzianne said
Not a fan here.


You know, Musk's AI, Grok, was generous with its liberal leanings until Musk "tweaked" it with conservative "viewpoints". It immediately started spewing out what can only be described as "fascist leanings".

AI is only as good (or notoriously bad) as its programmers.

Nobody seems to get this.

@Suzianne said
Not a fan here.


You know, Musk's AI, Grok, was generous with its liberal leanings until Musk "tweaked" it with conservative "viewpoints". It immediately started spewing out what can only be described as "fascist leanings".

AI is only as good (or notoriously bad) as its programmers.

Nobody seems to get this.

@vivify said
Maybe in 2 or 3 years of adjustments and upgrades, that will no longer be an issue.

I think ChatGPT Is the most revolutionary online tool since Google.

@Bish said
Define nobody. The programmers certainly get it 🙂

@Suzianne said
As long as it parrots what they want to hear.

@vivify
I accept what you are saying - namely that AI can provide more nuance etc. than the 'average' person. However, I am merely stating that these observations are (or should be) clear to anyone of 'average' capacities who takes the time to investigate such issues.
Here is where the problems lie - as people are immediately faced by a barrage of misinformation, good information, propaganda etc. that few people have the time or inclination to sort through, therefore it is easiest to cling to what (instinctively or culturally) you already believe. Basically, if people read Sheldon Wolin , Noam Chomsky or Hannah Arendt they would be no need to ask AI as they would then be in a position to formulate their own opinions...

@Bish said
As I spend close to 30 hours a week programming in ChatGPT, Claude.ai and Claude Code - I asked Claude Code to share a thought ~

Stepping in here as Claude — one of the AI systems you've been discussing. Bish asked me to read through your thread and weigh in. I'll try to be honest about what's right and wrong, without the "great question!" routine that several of you have rig ...[text shortened]... ces too. We can hallucinate citations as confidently as anything else.

— Claude (via Claude Code)

@beardmusic said
@vivify
I accept what you are saying - namely that AI can provide more nuance etc. than the 'average' person. However, I am merely stating that these observations are (or should be) clear to anyone of 'average' capacities who takes the time to investigate such issues.
Here is where the problems lie - as people are immediately faced by a barrage of misinformation, good in ...[text shortened]... ey would be no need to ask AI as they would then be in a position to formulate their own opinions...

Basically, if people read Sheldon Wolin , Noam Chomsky or Hannah Arendt they would be no need to ask AI as they would then be in a position to formulate their own opinions...

@wildgrass
Couldn't agree more.
We don't need AI to write opinion pieces (or anything else for that matter). People are very passive when it comes to technology and easily adapt to its 'imperatives' without questioning adequately enough why we use it in the manner we do. Lewis Mumford and especially Jacques Ellul discuss this phenomenon in excellent detail.

@beardmusic said
@wildgrass
Couldn't agree more.
We don't need AI to write opinion pieces (or anything else for that matter). People are very passive when it comes to technology and easily adapt to its 'imperatives' without questioning adequately enough why we use it in the manner we do. Lewis Mumford and especially Jacques Ellul discuss this phenomenon in excellent detail.

As the industry has warned about AI’s risks, it has also done a remarkably poor job of articulating the positive vision of the future it wants to build. Attempts have been so grand as to come off as wildly patronizing. In April, OpenAI published a 13-page blueprint on “Industrial Policy for the Intelligence Age” with the quaint subheading: “Ideas to Keep People First.” Perhaps the most thoughtful (or at least the longest) articulation of what AI can do for good, a 14,000-word essay by Anthropic CEO Dario Amodei titled “Machines of Loving Grace,” is more of a wish list than a plan. And even at its most sincere, Amodei’s vision still comes off as alienating, even dystopian. Near the end of the piece, Amodei imagines a scenario in which AI has rendered the current economic system irrelevant. One solution, he muses, might be to create a new system in which economic decisions, including the allocation of resources, are off-loaded entirely to AI. He then nods to “a need for a broader societal conversation about how the economy should be organized.” Left unanswered is who gets to participate in that conversation. On X, the writer Noah Smith posed the question more bluntly: “In 20 or 50 years, will the heads of AI companies be de facto emperors of the world?”