AI General Thread

[the-pi-guy]

You'd think a bunch of "programmers" on reddit would understand large language models better. Reddit - Dive into anything

This simpson's pic with text is fake yet so many people think it's a database that copy and pastes answers.

I don't know if the "programmers" understand anything about computers...

2023 and windows STILL doesn't know what a .cpp file is.

[Legend]

It almost feels like a coping mechanism with how the general public thinks of ai as a copy and paste machine. From reddit to artists to tv shows, that take exploded in popularity.

I wonder what it will take to change that perception, maybe permanent ai memory? The current submit query => get query setup is pretty limited.

[kitler53]

https://arstechnica.com/information-technology/2023/04/surprising-things-happen-when-you-put-25-ai-agents-together-in-an-rpg-town/

[Legend]

https://arstechnica.com/information-technology/2023/04/surprising-things-happen-when-you-put-25-ai-agents-together-in-an-rpg-town/

Can't wait for this to be done in real life with robots. Westworld here we come!

[kitler53]

my company just banned/forbid the use of chatGPT on company hardware/information:

i hadn't heard about this before but they pointed to several concerns one of which was this:
Three Samsung employees reportedly leaked sensitive data to ChatGPT

[Legend]

my company just banned/forbid the use of chatGPT on company hardware/information:

i hadn't heard about this before but they pointed to several concerns one of which was this:
Three Samsung employees reportedly leaked sensitive data to ChatGPT

Makes sense. OpenAI is working towards being secure and safe for private data, but I definitely wouldn't trust it yet.

[Legend]

I decided to try recreating Square Enix's tech demo in chatpgt. Worked better on the first try.




(the square enix tech demo has more commands so at worse you'd need GPT4 to do better)

[Legend]

Fast food restaurants are trying out AI for drive thru ordering. The example I saw on the news sucked but with current tech it should be nearly perfect.

Wonder if this will quickly become standard or if like touch screen ordering it'll be in and out a few times.

[kitler53]

drive through has yet to master the tech of "a speaker that sounds clear".  I have no confidence that can get AI ordering to function. 

it will be like my Alexa:

me "Alexa - play foo fighters"

Alexa:  "the capital of Minnesota is st. Paul". 

[Legend]

drive through has yet to master the tech of "a speaker that sounds clear".  I have no confidence that can get AI ordering to function.

it will be like my Alexa:

me "Alexa - play foo fighters"

Alexa:  "the capital of Minnesota is st. Paul".

I can't wait for alexa to get large language model ai. She hears me right but has the stupidest output most of the time.

[kitler53]

I can't wait for alexa to get large language model ai. She hears me right but has the stupidest output most of the time.

the one that gets me the most is:

"alexa, turn up"  ...alexa turns off.

i have to always remember to thrown the word "volume" in or it fails 100% of the time.

[the-pi-guy]

Like as close-minded as we were with NFT, blockchain and crypto currencies?

I'm glad you bring that up, because its an awful argument. The opponents of all of the above (myself included) point to their failings as to why they are bad. The opponents of AI, point to it's success and dangerous power to actually succeed in what it promises, as to why it's so bad. It''s fundamentally the exact opposite position. NFTs et al weren't (and aren't) the future because they're dogshit scams propped up by morons. AI is the future, despite people wishing to "ban it", whatever that looks like.

I thought this was an interesting take from Era.

[Legend]

I see very few opponents of AI acting that way. Instead it's always about acting as if the tech sucks and pointing out its failures.

[kitler53]

it's not just AI,.. a lot of supposed tech enhancements suck.   autocorrect typically does more harm than good.  speech to text is awful any time it's not something like a call center where a very specific question is asked and all it needs to do if distinquish "yes" from "no".   i literally think that the way google/facebook/ect only show you "results you'll be interested in" is a primary factor in the polarization of america and the eventual downfall of demacracy.  clearly these tools have been put to far greater use by authoritarians like russia/china then it has in free countries.

i can see a few ways in which AI could make lives better.  i just also see the far more ways that AI will be used inappropriately to the detriment of humanity.   i promise you it's not going to be long before someone utilizes this tech in cars/drones as murder bots.   it only takes 1 bumb in the world and I assure you there are far more than 1.  has horrible as nukes are at least they aren't easily accessible.

[the-pi-guy]

I see very few opponents of AI acting that way. Instead it's always about acting as if the tech sucks and pointing out its failures.

I've mostly seen a lot of concern on Era about AI pushing people out of jobs.