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

51–60 of 652 posts

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

#51

Post by dave4420 »

lazide wrote: post #46
K0balt wrote: post #35

There is a bit of very important content missing from the explanation of the autocomplete analogy. The combination of encoding / tokenization of meanings and ideas, related concepts, and mapping these relationships in vector space makes LLMs not so much glorified text prediction engines as browsers/oracles of the sum total of cultural-linguistic knowledge as captured in the training corpus. Understanding how the impl…

It’s also why they can produce such hard to identify bullshit and harmful output. I’ve had some really convincing, yet fundamentally flawed, code output that if I hadn’t done about a million code reviews before I might have just used. And been totally screwed later. Near as I can tell, that the bullshit is so much more convincing with them is a huge detriment that society really won’t learn to appreciate until it’s g…

The uncanny valley of competence.

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

#52

Post by iamacyborg »

KronisLV wrote: post #11

This is a pretty admirable goal! I'm saying this unironically, but I wish there were courses on looking at information critically and more in how to have a healthy and safe life in the modern day world (including things like data security, how to deal with social media etc.) that would be taught to everyone in schools/colleges/universities. In my country, there are still public announcements about not trusting random…

People frequently laugh about it, but media studies goes into what you describe.

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

#53

Post by vladms »

einrealist wrote: post #49

This website is so important! Now ask yourself why AI companies don't want to be regulated or scrutinized. So many companies (users and providers) jump on the AI hype train because of FOMO. The end result might be just as destructive as this mythical "AGI". Edit: I am not saying to not use the technology. I am just on the side of caution and constant validation. The technology has to serve society. But I fear this hy…

My impression is that companies in most of the fields do not like to be regulated or scrutinized, so nothing new there.

While observing some people using LLMs, I realized that for a lot of people it really makes a huge difference in time saved. For me the difference is not significant, but I am generally solving complex problems, not writing nicely formatted reports where words and not numbers are relevant, so YMMV.

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

#54

Post by iamacyborg »

picafrost wrote: post #20

I think a great number of working professionals need a course like this too. I am already tired of ChatGPT being cited by the less experienced as an invisible expert in the room during technical discussions.

This stuff is so frustrating, I have colleagues who sent long, clearly AI generated documents who don’t seem to understand that if they can’t be arsed to write something, why should I bother reading it?

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

#55

Post by IceDane »

aqueueaqueue wrote: post #31
joshdavham wrote: post #16

Do you feel that you may be being a bit provocative by calling LLM's 'bullshit machines'? I understand the frustration as I've been bullshitted by these models just as much as the next programmer, but surely with recent advancements in RAG and reasoning, they're not just 'bullshit machines' at this point, are they?

I have been thinking about this and I have an idea for local LLM I want to try. It is based on the assumption LLMs make mistakes (they will!) but be confident about it. Use a 1B model and ask it about a city for fun mistakes in an otherwise impressive array of facts. Then you will see how bullshitty LLMs are. Anyway the idea is to constrain the LLM and a language understanding and to choose a constrained response. Fo…

This is basically how a lot of it works already. What you are describing is more or less just structured output and tool usage.

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

#56

Post by aqueueaqueue »

IceDane wrote: post #55
aqueueaqueue wrote: post #31

Earlier quoted context omitted.

I have been thinking about this and I have an idea for local LLM I want to try. It is based on the assumption LLMs make mistakes (they will!) but be confident about it. Use a 1B model and ask it about a city for fun mistakes in an otherwise impressive array of facts. Then you will see how bullshitty LLMs are. Anyway the idea is to constrain the LLM and a language understanding and to choose a constrained response. Fo…

This is basically how a lot of it works already. What you are describing is more or less just structured output and tool usage.

Yes! Thanks for the terms. I guess I am saying restrict it to that. Probably how Siri etc. works. This gives a low bullshit usage pattern.

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

#57

Post by ta8645 »

CyMonk wrote: post #44
joshdavham wrote: post #16

Do you feel that you may be being a bit provocative by calling LLM's 'bullshit machines'? I understand the frustration as I've been bullshitted by these models just as much as the next programmer, but surely with recent advancements in RAG and reasoning, they're not just 'bullshit machines' at this point, are they?

they define the term bullshit in lesson 2: [quote]BULLSHIT involves language or other forms of communication intended to appear authoritative or persuasive without regard to its actual truth or logical consistency.[/quote]

That is an emotionally manipulative definition and anthropomorphizes the LLMs. They don't "intend" anything, they're not trying to trick you, or sound persuasive.

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

#58

Post by KronisLV »

XorNot wrote: post #21
KronisLV wrote: post #11

This is a pretty admirable goal! I'm saying this unironically, but I wish there were courses on looking at information critically and more in how to have a healthy and safe life in the modern day world (including things like data security, how to deal with social media etc.) that would be taught to everyone in schools/colleges/universities. In my country, there are still public announcements about not trusting random…

Most scam prevention fails though because the world is full of exceptions. Like it's mind-blowing to me that charities still call you and ask for credit card details over the phone and this is like...a legitimate way to go about things. Or that any government agency calls you and doesn't just leave a verifiable number to call the operator back on.

> Like it's mind-blowing to me that charities still call you and ask for credit card details over the phone and this is like...a legitimate way to go about things.

> Or that any government agency calls you and doesn't just leave a verifiable number to call the operator back on.

That's rather unfortunate! I wonder if in those cases it'd be better to tell them that you'll get in contact through e-mail or something, because then at least it's you going to their actual homepage, looking up contact details and communicating through that.

In my country, we also have a bunch of governmental e-services, one of which is a web based communication platform with most institutionns (translated description, because they haven't bothered to translate it themselves, and also sometimes block connections from outside the country):

> An e-address, or official electronic address, is a personalized mailbox on the Latvija.gov.lv portal for unified and secure communication with state and local government institutions. The e-address system organizes secure, efficient and high-quality e-communication and e-document circulation between state institutions and private individuals, ensuring data confidentiality and protection of personal data from unauthorized access, unlawful processing or disclosure, accidental loss, alteration or destruction. An e-address is not e-mail, but its use is similar. Communication in an e-address is confidential, and the data is guaranteed to be available only to you and the institution you contacted. The main purpose of an e-address is to replace registered paper letters with electronic ones in cases where a state administration institution needs to send information and documents to a specific resident or entrepreneur. Citizens and entrepreneurs can also contact more than 3,000 institutions at any time and from any location via E-address. These include not only state and local government institutions, such as the Food and Veterinary Service, the State Labor Inspectorate, the Competition Council, etc., but also judicial institutions, sworn bailiffs and insolvency administrators, as well as private individuals to whom state administration tasks have been delegated.

That seems like a pretty good common sense idea for organizing trusted 2 way communication.

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

#59

Post by tialaramex »

This course mentions the famous Apple advertisement. Unfortunately it slightly oversells and while I'm sure that's not because this fragment was written by an LLM it is exactly the sort of over-simplification which leads to LLMs generating wild bullshit when they interpolate this "fact" with other "facts" they've been fed, and we ought to strive to do better when writing for humans.

"Describe how prior to 1984, there was no such thing as a graphical user interface, visual desktop, an intuitive menu system, or mouse-based navigation."

Apple were offering a mass market product which had these features so that's important - but there had been "such a thing" for quite some time before that. Douglas Engelbart's "Mother Of All Demos" in 1968 -- Sixteen years earlier shows all the features you mentioned. https://en.wikipedia.org/wiki/The_Mother_of_All_Demos

Unfortunately the demo is very long for a modern audience, so unlike "Watch a Superbowl ad" it's a hard sell to show the entire demo, but do go watch for yourself.

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

#60

Post by Almondsetat »

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 are badly produced (clipping audio, interviewer doesn't have a dedicated microphone) and it's not even clear why they're there.

Please, take inspiration from actual e-learning platforms.

Post reply on HN