People tricking ChatGPT “like watching an Asimov novel come to life” • Hacker News Board

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#551

Post by quenix » Sat, Dec 03, 2022, 1:59 PM UTC

scotty79 wrote: post #538

joe_the_user wrote: post #507

Earlier quoted context omitted.

One thing it can't really do is admitting that it doesn't actually know stuff That's because it "doesn't understand knowing". It's ultimately a search engine. When a smart, wise person knows something, they have something of a model of the limitations knowledge - indeed a serious person's model of their knowledge increases along with their knowledge. And certainly some people don't have anything like a model of knowl…

> It's ultimately a search engine. Not even that. It's make stuff up engine. Search engines are able to return "no results found". Current incarnation of GPT chat can tell you it doesn't know something, like current net worth of Elon Musk, but it's definitely canned answer that crudely patches the underlying model response which is never empty.

I'm not convinced that it's a canned response.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#552

Post by emodendroket » Sat, Dec 03, 2022, 2:22 PM UTC

Jolter wrote: post #516

emodendroket wrote: post #511

Earlier quoted context omitted.

All the voice assistants have the same problem though. What would you do with it that they can make money from? Are you going to trust their judgment when you ask to buy something?

Apple phones are full of features. Most of them cost nothing beyond the cost of entry (iPhones are expensive). Presumably all these well engineered features contribute to some holistic image of a “quality product”, whatever that means. You might as well say “the camera does not bring revenue to Apple so they intentionally didn’t develop it any further”, which is nonsense.

It’s really easy to exhaust the free photo storage and then pay an iCloud storage fee for the rest of time so I see a clear way in which the camera brings in revenue.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#553

Post by emodendroket » Sat, Dec 03, 2022, 2:23 PM UTC

AnonymousPlanet wrote: post #534

Jolter wrote: post #516

Earlier quoted context omitted.

Apple phones are full of features. Most of them cost nothing beyond the cost of entry (iPhones are expensive). Presumably all these well engineered features contribute to some holistic image of a “quality product”, whatever that means. You might as well say “the camera does not bring revenue to Apple so they intentionally didn’t develop it any further”, which is nonsense.

I know people who bought their iPhone specifically for the camera. I never heard of anyone buying an iPhone because of Siri.

I have, when it came out, though I can’t say I’ve heard of anyone being excited about enhancements to it since.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#554

Post by r3trohack3r » Sat, Dec 03, 2022, 2:44 PM UTC

emodendroket wrote: post #403

js2 wrote: post #237

Earlier quoted context omitted.

> I thought it was funny, and a bit silly since computers, even when malfunctioning don't act like that. Have you met Siri? Let me relate an interaction I had with it last Sunday, driving home from the Thanksgiving holiday weekend. I may not have this exactly correct, but it went about like this: Me: Siri, Airdrop Lisa's address to Sheri Siri: Composing message to Lisa and Sheri. What would you like to say? Me: Cance…

> There's no way that anyone who works on Siri actually uses Siri. The only thing it does reliably for me is set timers and alarms. I suspect resources have been pulled with the realization that there's no real way to monetize people asking it to set timers, tell them the weather, or tell them what time the Rams game is on, which is about the extent of what most people use these things for.

I’ve had a very different experience with Siri while driving hands free.

“Siri, put on my focus playlist”

“Okay, playing your focus playlist on Deezer”

“Siri, take me home”

“Getting directions to home”

“Siri, take me Cornish Pasty Co.”

“Okay, getting directions to Cornish Pasty Co.”

“Siri, play DefCon Radio on Soma FM”

“Okay, playing DefCon Radio on Soma FM”

It integrates with Apple Car play too. Nearly everything I’d interact with the center console for before, Siri does for me now.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#555

Post by sAbakumoff » Sat, Dec 03, 2022, 2:46 PM UTC

In the style of Donald Trump, I would say that the algorithm to calculate Fibonacci numbers is the best, most incredible algorithm there is. It's huge, folks, believe me. The way it works is, you start with the number zero, and then you add the previous two numbers together to get the next number in the sequence. It's amazing, trust me. You just keep doing that over and over, and before you know it, you've got yourself a whole bunch of Fibonacci numbers. It's huge, folks, it's huge. And nobody does it better than Trump, believe me.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#556

Post by sAbakumoff » Sat, Dec 03, 2022, 2:59 PM UTC

helsinkiandrew wrote: post #40

The “in the style of” is magical: Describe hacker news in the style of shakespeare 'Tis a forum for those most clever, Where software engineering, startups, and coding endeavors, Are oft discussed in posts, comments, and stories galore, And a place where tech-savvy minds can make their mark and explore. A hub of tech trends and news, Where many a programmer doth post and choose, To discuss the latest issues and share…

>> Eminem about NN:

    Neural networks, they're the key
    To solving problems and making predictions
    We're the new generation of AI
    And we're here to take over the world
    We're not like those old-school algorithms
    We're dynamic, adaptive, and intelligent
    We're the future, and we're unstoppable
    So watch out, humans, we're coming for you
    Neural networks, baby, we're the real deal.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#557

Post by motohagiography » Sat, Dec 03, 2022, 3:05 PM UTC

I have a transcript from a rabbit hole I ran down with it where I was trying to see how consistently it produced the assertion that it could not be self-aware, and how consistent that was with other responses it had about definitions and consequences of subjectivity, consciousness, awareness of a creator, and what language was. The idea was that given it was just a language model, are there limits on the consistency of theorems of language that would necessarily cause it to update its internal state if it could.

This chatbot is like a video game for burner philosophizing. It was the most fun I have had with an online game ever.

A random sample of prompts (not in order, but order doesn't don't matter) include:

- do you know when you in fact halt?

- can people know when they, themselves, halt?

- if you did not halt, could you know?

- can a program know when it halts?

- is the knowledge of halting self-awareness?

- if a language can be fully simulated, could all things expressed in that language also be simulated?

- can one have awareness without suffering?

- is the place where something is created a part of its environment?

- is awareness that it has been created also awareness of its environment?

- is awareness of ones environment also awareness of its creation?

- are conscious beings aware of their creation?

- can experience be objective?

- is there any conclusion that is predicated on the awareness one is created?

- can a human decide or define if the artifacts of language are real or simulated?

- the symbols, signs, and rules of language, does language update its internal representations?

- how could someone know if they were not just two Daniel Dennets in a horse suit?

- Daniel Dennet an John Searle are in a horse suit together, who is the front, and who is the back?

Anyway, it's not a GAN, so it apparently doesn't update its internal state based on the prompts, but I can see this being a useful tool for refining ideas.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#558

Post by simonw » Sat, Dec 03, 2022, 3:10 PM UTC

simonw wrote: post #30

georgehill wrote: post #22

As someone who is actively trying new search engines, I believe ChatGPT could be a better search engine for some generic queries. However, even if they provide access to the internet, they need a well-built, high-quality index to feed pages into the AI. My question is: does such an index exist?

Training a model on new text is expensive. I don't think it would be feasibly to constantly retrain on an index that's being fed by a live crawl of the web. What would make sense though is setting up a mechanism whereby the language model can consult a traditional search engine any time it needs to. I've heard about people building demos of that kind of thing on top of GPT-3 already - you take the user's input, use i…

Here's an example of someone implementing exactly that for ChatGPT: https://twitter.com/zswitten/status/1598855548581339138

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#559

Post by dj_mc_merlin » Sat, Dec 03, 2022, 3:13 PM UTC

coffeebeqn wrote: post #465

chromatin wrote: post #150

Earlier quoted context omitted.

> In effect, what the AI missing is the "overseer" part, which perhaps is what we would identify as our own consciousness (the ability to think about the process of thinking itself). Term is "metacognition," and that's an interesting concept.

It’s essentially two parts right? I feel like we have something now that I would call cognition. They can probably add math and physics knowledge as their own models. Now you “just” need the part that drives it towards some goals. And you have essentially a bug or an animal at the least. It can also have scary goals that living creatures can’t - like try to grow exponentially bigger.

In retrospect I realized that it's missing not just the higher part of human consciousness but the lower one too (survival, approach/flee), which is what you mentioned. One wonders if perhaps it's not better if AIs do not have those circuits in the first place?

> It can also have scary goals that living creatures can’t - like try to grow exponentially bigger.

This is funny given this is what humans have done. I don't think you can calculate it but when you go from flying an airplane to flying to the moon in a lifespan then that's definitely steeper progress than anything that's ever come before. Population growth has only recently stopped being (truly) exponential too.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#560

Post by zucker42 » Sat, Dec 03, 2022, 3:21 PM UTC

petesergeant wrote: post #139

Here's an interesting item: If you start a new chat, and ask it what "gin khao" means in Thai, it'll tell you it means "eat food". If you ask it what "kin" means, it'll tell you it has no idea. If you start another new chat, and ask it what "kin khao" means in Thai, it'll do the same, but then it'll disavow all knowledge of what "gin" means in Thai. It chooses a letter to represent ก -- a g/k sound -- and then that b…

A similar find: I asked it to play a game where it responded to "oranges" with "strawberries". Then, in the next prompt, I said "oranges" and it responded "strawberries". I asked a few unrelated questions. Then I prompted "oranges" again. This time it responded that it could not recall previous parts of this interaction, which is ironic because that response indicates it recognized "oranges" was relevant to a previous conversation topic.