So "X" - nothing to see here. Elon's in control - Part XXX **Threadbans in OP**

AndrewJRenko

https://www.boards.ie/discussion/comment/120846095#Comment_120846095

Try it with Obama

Quin_Dub

https://www.boards.ie/discussion/comment/120846095#Comment_120846095

Works for me...

https://www.boards.ie/discussion/comment/120846198#Comment_120846198

Right, but it's still a major problem with AI. When companies' reputations are at risk, they spent millions training a safe bias into them.

I've played around with uncensored open source ones based on Llama and it's pretty wild when you can get them to say anything.

pixelburp

https://www.boards.ie/discussion/comment/120846095#Comment_120846095

https://www.boards.ie/discussion/comment/120846198#Comment_120846198

Worked here for me too; it's even vaguely coherent and pertinent, bar the last line which is just nonsense. I've never been on a TV show, only radio. So not sure what to make of your own response, why you got pretty aggressive guard-rails?

Anyway: to create a truly creative AI, one that follows its own patterns and ingested input, without external stimulus or bias? Good luck there to Elon Musk 'cos I'm not sure how you either create or prove that 'cos then you're entering the realm of the Turing Test; while many a Sci-Fi writer worth their salt has tried to parse the idea of robotic/digital sapience - and how one tells the difference between programming & "free will". Did the AI tell that racist joke 'cos it's truly bias-free, or that it (or its maker) just a racist prick?

Point being: you might say "let's build an AI without bias", but that feels utterly impossible; bordering on the boast to make a perpetual motion machine. Not if the AI is built by humans, especially ones with hardened biases of their own. But what I know about the technology behind Machine Learning and beyond I culd fill on a postage stamp, so we'll see.

Quin_Dub

https://www.boards.ie/discussion/comment/120846495#Comment_120846495

Not altogether sure it's a "problem" let alone a major one.

Of course it's been programmed to not expose the owners to libel - why would that be a "problem" ??

timetogo1

https://www.boards.ie/discussion/comment/120846495#Comment_120846495

But Musks comment was about a "liberal" bias. Does liberal bias = safe bias?

Quin_Dub

https://www.boards.ie/discussion/comment/120846610#Comment_120846610

"Liberal Bias" often means "Won't let me be a dick" when people like Musk say it.

https://www.boards.ie/discussion/comment/120846562#Comment_120846562

This isn't thread for it but a lot of academics and politicians understand that this is a problem with AI. It seems you think it's just some small annoyance for people you perceive to be right-wing racists or something.

If China released a superior AI some day, or one that was far cheaper or something, you probably wouldn't be so dismissive of bias if it was incorporated inside a lot of your daily life.

Christy42

https://www.boards.ie/discussion/comment/120846676#Comment_120846676

I mean the US had Musk on the stand in front of politicians which just showed they didn't have a clue how Facebook worked never mind AI so I am not trusting any politician on that.

But yeah presumably chatgpt is biased though if you want to make the point a quickly disproved example is not going to help. If the team gathered by Musk makes an Ai that will also be biased. Generally we should not involve chatgpt in moral decisions nor one developed by China or xai.

Better than xai Musk would be better of investing in education to teach people about it's limits rather than just making a different biased AI. Assuming his motivation is actually that.

Quin_Dub

https://www.boards.ie/discussion/comment/120846676#Comment_120846676

Like you say not the thread for it, but as Pixelburp says all AI will have a bias - It's completely and utterly unavoidable.

And given the general mix of available information globally , that bias will be of a more liberal (from a US perspective) viewpoint as that is where the majority of people are so that is what the content being ingested will reflect.

timetogo1

As for bias, it can be manipulated so maybe it depends on the prompts you had entered previously.

pixelburp

Seems a reasonable guess that an AI tool might also use each user's previous prompts when fashioning a result; that the AI would in turn try to figure out what results a given user might like, as well as want. It'd be a good test of the AI's flexability. God knows it's the main driver of the algorithms on social media these days. For context: my Joe Biden limerick was my first GPT usage as I signed up out of curiousity from this conversation.

https://www.boards.ie/discussion/comment/120846753#Comment_120846753

I didn't have anything entered previously.

https://chat.openai.com/share/50bb1a79-68b4-4540-8e0e-a9f3e440d9a7

https://www.boards.ie/discussion/comment/120846778#Comment_120846778

It famously doesn't take that into account. I've gotten multiple Joe Biden limericks since that first failure. I'd love to know what your point here could be.

pixelburp

https://www.boards.ie/discussion/comment/120846803#Comment_120846803

I'm trying to spitball why OpenAI would be rather inconsistent and random when it comes to the prompt we've been trying; there's clearly something at play that the same prompt yields limericks sometimes and not others. You say "famously" but I've said already there's a ceiling of my knowledge of GPT. Relax, I'm not hinting or snarking at anything about you.

https://www.boards.ie/discussion/comment/120846817#Comment_120846817

It works sometimes and it doesn't work other times because everything the AI does is random but based on weights. Here is one where the prompt worked where I tested if it had a problem doing two in a row.

On a side note, I've never experienced ChatGPT running so badly. Failed loading of histories and failed attempts at sharing.

Brussels Sprout

I think that it's quite telling that Musk's go-to BS excuse is to do with "bots". He used it as an excuse to try and weasel out of the stupid price that he offered for the company. It was one of the reasons he gave for Twitter Blue and he brought it up again during the recent outage. Nonetheless under his stewardship the number of bots has dramatically increased.

I now get notifications like this:

and I have noticed that if a big football match is happening and I search by the hashtag that I often have to scroll past a bunch of bot-generated (but with a blue check-mark of course) tweets promoting "Pepe Coin" or some other nonsense to actually see any real tweets.

Penn

I can't help but wonder if there are a bunch of ChatGPT engineers in a room somewhere looking at data from today and wondering "What's with all these limericks...."

nachouser

When it eventually goes full Skynet, you now know the first county Chatty will target.

Quin_Dub

https://www.boards.ie/discussion/comment/120846919#Comment_120846919

Along with "Who or what the hell is a Pixelburp??"

pjohnson

https://www.boards.ie/discussion/comment/120846940#Comment_120846940

Pixelburp defeated Clare by a Joe Biden in the All Limerick Hurling Championship.

pixelburp

Thanks to you lot I'll be AI's first victim when Musk's sociopathic, unbiased version gains sapience and goes rogue.

Penn

https://www.boards.ie/discussion/comment/120846972#Comment_120846972

In fairness, if it's Musk's AI it probably won't kill you unless you subscribe to AI-Killbot Blue, and even then it won't work right.

nachouser

https://www.boards.ie/discussion/comment/120846972#Comment_120846972

Only if you're trans.

kerplun k

This thread starting to reach peak entertainment. I want even more rhymes that include various politicians and pixelburp.

Cordell

https://en.wikipedia.org/wiki/Roko%27s_basilisk

Do your part.

silliussoddius

Sounds like something someone made up after reading “I have no mouth and I must scream”.

notnattynate

I see a lot of happy tweeters posting earnings for their content on Twitter last couple of hours.

I thought twitter was going broke?

Quin_Dub

https://www.boards.ie/discussion/comment/120848947#Comment_120848947

Twitter ARE going broke , or at least they are losing money hand over fist - Broke is a relative term given Musks billions.

It's more that this is the first time that Twitter have ever paid any content creators for posting on the platform so people are sharing what they are making.

Very small potatoes compared to earnings from YouTube etc. by comparison but they are getting paid at least.

Penn

It's also as much a sales tactic as anything else. Subscribe to Twitter Blue and you can earn some of that money back. Some of the bigger accounts will earn way more than the subscription. But the vast majority of Twitter Blue users, the ones already complaining that even Twitter Blue pushing their tweets to the top hasn't led to any greater engagement/followers, will get a pittance. But the more people who subscribe to Twitter Blue thinking they can earn money back (particularly when you have some of the biggest Musk supporters showing how much they're getting) the more money goes directly to Twitter.