Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

thebullshitmachines.com

Post reply on HN
Print view View original

261–270 of 652 posts

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#261

Post by _dark_matter_ »

BugsJustFindMe wrote: post #252

> Large language models are both powerful tools, and mindless—even dangerous—bullshit machines. Yes, but so is the average person. Do you cover the similarities/parallels to common human behavior patterns in your course? It stands out to me as a major blind spot in a lot of the discourse about LLMs, down to willing acceptance of humans intuiting what someone else meant when the words themselves were fundamentally amb…

The key distinctions are: we can hold people accountable, and the amount of shit produced by one person is limited. Neither of those are true for LLMs.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#262

Post by I-M-S »

jvanderbot wrote: post #258

> We (the authors of this website) have at times sought insight into the inner workings of an LLM by asking it “why did you just do that?” > But the LLM can’t tell us. It’s not a person. It doesn’t have the metacognitive abilities necessary to reflect on its past actions and report the motivations underlying them*. > With no clue why it did whatever it just did, the LLM is forced to guess wildly at a plausible explan…

Related, since the advent of LLMs I've become acutely aware how any argument of a considerable length with another person quickly starts meandering and how topics change seemingly of no one's volition - almost as if our own internal token limit has been exceeded.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#263

Post by BugsJustFindMe »

justonceokay wrote: post #256
BugsJustFindMe wrote: post #252

> Large language models are both powerful tools, and mindless—even dangerous—bullshit machines. Yes, but so is the average person. Do you cover the similarities/parallels to common human behavior patterns in your course? It stands out to me as a major blind spot in a lot of the discourse about LLMs, down to willing acceptance of humans intuiting what someone else meant when the words themselves were fundamentally amb…

Yes but people have culpability and responsibility. In places of power or influence, this culpability can lead to being fired, legal action, disbarment, loss of money, etc. so there is a real pressure to be coherent and aligned with reality.

You say this but I have a ton of experience that "can lead" and "there is" come with a gigantic pile of caveats to the extent that they appear to be more false than true. Or they're technically true but practically meaningless. The world has been utterly awash in mass-perpetuated misinformation for at least all of recorded history without any real ability to stem the onslaught. This is not a modern problem just because a modern technology also exhibits it.

You should look up the percentage of Americans who believe in ghosts sometime. About as many people believe in ghosts as don't, so no matter which side you land on, the other side is enormous. One of the sides must be wrong, I won't claim which, though only the belief side fails a falsifiability check. Where's our accountability to believing and spreading ideas based in reality again? The believers believe because they learned about it from someone. It didn't happen spontaneously on its own.

It's all just been memes the entire time.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#264

Post by BugsJustFindMe »

_dark_matter_ wrote: post #261
BugsJustFindMe wrote: post #252

> Large language models are both powerful tools, and mindless—even dangerous—bullshit machines. Yes, but so is the average person. Do you cover the similarities/parallels to common human behavior patterns in your course? It stands out to me as a major blind spot in a lot of the discourse about LLMs, down to willing acceptance of humans intuiting what someone else meant when the words themselves were fundamentally amb…

The key distinctions are: we can hold people accountable, and the amount of shit produced by one person is limited. Neither of those are true for LLMs.

You wrote basically the same thing as the adjacent person, so rather than write the same response to you both, I'll redirect to my reply over there:

https://news.ycombinator.com/item?id=42993981

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#265

Post by allenu »

Almondsetat wrote: post #60

I'm sorry, but this website is awful. Not only does it have an illogical structure (table of contents at the end? no "next lesson" button? gigantic images that fill the entire screen?), but the aesthetic of the entire thing is off. It tries to be sleek and modern with scrolling animations, but they are janky and rigid and the images are rectangles put in front of a bad gradient. Not to mention the video interviews ar…

Totally agree here. I visited the page and scrolled through it to see what it was all about and saw a bunch of pull quotes and couldn't work out what I was looking at. It just looks like a light magazine article or brochure until you either click on the hamburger menu to see a full table of contents or scroll to the very bottom to see the table of contents there.

This really needs some design improvements if they want people to read through to the actual lessons. Most people are going to drop off after scrolling through that landing page.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#266

Post by therein »

idunnoman1222 wrote: post #257

My 80s copy of the encyclopedia Britannica was riddled with errors, perhaps we will survive this post truth

The situation is closer to if we had 10,000 variants of Encyclopedia Britannica in 80s that all looked like distinct bodies of work, riddled with different errors while looking like they were written from scratch.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#267

Post by 1propionyl »

_dark_matter_ wrote: post #261
BugsJustFindMe wrote: post #252

> Large language models are both powerful tools, and mindless—even dangerous—bullshit machines. Yes, but so is the average person. Do you cover the similarities/parallels to common human behavior patterns in your course? It stands out to me as a major blind spot in a lot of the discourse about LLMs, down to willing acceptance of humans intuiting what someone else meant when the words themselves were fundamentally amb…

The key distinctions are: we can hold people accountable, and the amount of shit produced by one person is limited. Neither of those are true for LLMs.

The recent book Unaccountability Machine (Dan Davies) made the rounds a while back on this site a few years back. A few years before that, Ted Chiang's essay in the New Yorker entitled "Will AI Become the New McKinsey?" did likewise.

There's a bright red through line here. I get the sense that the intellectual ferment is starting to develop an awareness of the risk (and if you're cynical, potential) of LLM deployments in business as a systematic strategy for absorbing accountability for decisions.

For the time being, it usually seems like there's someone who is accountable, or at least can be scapegoated. But how long will that last? As Davies points out in his book, we didn't need LLMs to create bureaucracies where the buck fails to stop anywhere and instead irretrievably slides between the tracks. As Chiang points out, the "efficiency maximization" of McKinsey served as a way for organizations to outsource accountability for major decisions to an entity very, very good at working backwards from desired outcomes while acting like they were just led to a fait accompli by the numbers.

[0] https://libro.fm/audiobooks/9781805220794-the-unaccountabili...

[1] https://www.newyorker.com/science/annals-of-artificial-intel...

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#268

Post by _heimdall »

JKCalhoun wrote: post #247
_heimdall wrote: post #235

Earlier quoted context omitted.

> only going to succeed if "we" want it — find value in it. I'll be pleasantly surprised if that's how it turns out. At least so far market dynamics haven't really been much of a driver for LLMs. Those with the money think its the next big thing and are pouring cash both into the LLMs themselves and any product that slaps a "powered by AI" sticker on the box. That's not to say people aren't also actively choosing to…

I see it too — but I'll remind both of us that these are still very, very early days.

For sure. And I definitely have a bias showing here towards not trusting the person in charge to be benevolent or to actually know what the "right" thing to do is in the long run.

Re: Modern-Day Oracles or Bullshit Machines? How to thrive in a ChatGPT world

#269

Post by fancyfredbot »

TeMPOraL wrote: post #250
fancyfredbot wrote: post #245

Earlier quoted context omitted.

There's obviously nothing wrong with learning by reading, but the way you tell whether what you read is true is by seeing whether or not it fits in with observation of reality. That's the reason we're no longer reading the books about phlogiston.

> the way you tell whether what you read is true is by seeing whether or not it fits in with observation of reality The only way any of us ever gets to see "whether or not it fits in with observation of reality" is to see if they get an A or F on the test asking it. Seriously. The "moons of Jupiter" question is the only one of the above one gets to connect to an observation independent of humans, and then if they did…

Independently checking the information you read in textbooks is very difficult for sure. But it's still how we decide what's true and what's not true. If tomorrow a new moon was somehow orbiting Jupiter we'd say the textbooks were wrong, we wouldn't say the moon isn't there.
Post reply on HN