RuneScape Forum Archive | A.I. Should we be worried?

Jan

2006

Rooh

Forum Moderator Posts: 40,586 Sapphire Posts by user Forum Profile RuneMetrics Profile

Geoffrey Hinton the" godfather of artificial intelligence" has quit his job at Google, with a warning about the growing dangers of AI. and says he now regrets his work.

He also said that although AI Bots are less intelligent at the moment, he predicts that they soon will be.

https://www.bbc.co.uk/news/world-us-canada-65452940

There's no doubting they're fascinating, it seems like every day there's a new story about them and they do seem really useful... when used properly.

I'll admit I don't know much about them as yet but I found an old quiz that I made up years ago and put it into one... and the answers came back immediately and were 100%.

A lot of customer support on websites seems to be handled by them too, but I've found them to be very limited - if your query is anything other than a predefined issue they don't seem to be that good.

The thought of them being used in cyber crime is pretty worrying =/

Has anyone else tried them out? Any interesting experiences to share?

Who's the cat that won't cop out when there's danger all about?

02-May-2023 09:32:22 - Last edited on 02-May-2023 09:43:28 by Rooh

Origin Nexus

Posts: 322 Silver Posts by user Forum Profile RuneMetrics Profile

Chat AI doesn't actually have any intelligence, any sentience, or any actual knowledge about anything at all, other than the probability of word placement in human language. They're just pattern prediction algorithms.

While they do mimic human language quite well, they don't actually know what they're saying. All they know is that through their model training, each word in a sentence has a probability rating of what word should come next, based on previous analysis of the human language as commonly used.

How the AI presents itself is entirely based on what input was given to the training model. If an AI is trained solely on works of Shakespeare, any question you pose to it will be result in a response that somewhat resembles the language used in the works of Shakespeare. However, because there isn't a large enough volume of input (Shakespeare's works are limited in size and numbers) the responses won't actually make too much sense, but, it will mimic the language used.

A neural network does nothing on it's own, and after training, it's only useful for predicting patterns and mimicking the input given in it's training model.

Now, if someone comes along and creates a subroutine that can consciously think, then there's a problem to be concerned about. But I'm not sure how possible that actually is.

02-May-2023 09:55:07

Jan

2003

Tuffty

Forum Moderator Posts: 152,160 Ruby Posts by user Forum Profile RuneMetrics Profile

I see AI as lazyness really.

Nothing worse then interacting with a script on a website click A, B, C etc and hope it gets where you need and the answer you seek.

Humans are far better but the robbing companies wants free labour.

Am I worried naaaa not in the least. I can't see it having much effect within the next 20-30 years.

If it does come out it better have an OFF switch. Know what I'm saying.....

Comprehensive Account Security

What do snowmen have for breakfast? Snowflakes!

02-May-2023 09:57:34 - Last edited on 02-May-2023 09:57:51 by Tuffty

Origin Nexus

Posts: 322 Silver Posts by user Forum Profile RuneMetrics Profile

Tuffty said :
I see AI as lazyness really. Is it lazy though? I mean, it's just a tool, like any other tool. Is using a drill to drive screws instead of doing by hand with a screwdriver lazy?

If you needed a picture of a raccoon driving a motorcycle, are you really going to wait around for one to drive by so you can take a picture? Of course not, that could take hours and you have a deadline, so you're just gonna ask an AI to draw one for you:

02-May-2023 10:06:13

Sep

2005

Megycal

Posts: 18,097 Opal Posts by user Forum Profile RuneMetrics Profile

We were travelling to see our son and due to traffic conditions were going to be late arriving at a restaurant ( part of a national chain ) we'd booked for lunch. We tried to telephone to let them know and were put through to a chat bot which couldn't understand what we were trying to do. When it eventually realised we wanted to change a booking it then asked for the telephone number we'd used to make the booking...we had used a laptop.. So we gave up and arrived a bit late.

02-May-2023 10:20:46

Jan

2006

Rooh

Forum Moderator Posts: 40,586 Sapphire Posts by user Forum Profile RuneMetrics Profile

Origin Nexus said :

If you needed a picture of a raccoon driving a motorcycle, are you really going to wait around for one to drive by so you can take a picture? Of course not, that could take hours and you have a deadline, so you're just gonna ask an AI to draw one for you:

That takes the fun out of things a bit, I love messing around in Photoshop and I'd happily spend time creating something like that.

It'd be interesting to create something then see if Ai can do a better job.

Who's the cat that won't cop out when there's danger all about?

02-May-2023 10:55:05

Averia Light

Posts: 28,508 Sapphire Posts by user Forum Profile RuneMetrics Profile

Chat bots are a bit different than AI - at least right now. Chat bots look for certain responses and respond using a set amount of responses. LLMs, like Chat GPT, do have data to work with, but it isn't limited to the responses someone creates and it does have the capability to evolve in real time.

Like most new technology, the power doesn't come from basically using it like an easy Google, it is going to come from when people start using it with existing technology,

And, I'd say it is going to be similar to when computers became economically available in mass. Computers changed a whole lot of things and made a lot of jobs obsolete, however, there were also a lot of jobs created, too. For instance cyber security is still absolutely going to be a thing as AI can be used on both sides and the never ending game of cat and mouse will likely still exist well into the future. I'd also imagine data integrity and auditing are still going to persist as it is probably a good idea to make sure the input and output are what a user would expect over time.

Of course, this is all speculation as I've been using AI as an easy Google, so I don't know the depths of its capabilities. Overall, though, if you are getting up there in age, you'll probably see more benefits, but if you aren't then adaptation is probably going to be needed. And I swear I'm not going to let her know all the pain I have known

02-May-2023 11:37:08 - Last edited on 02-May-2023 11:48:57 by Averia Light

Feb

2005

Joel

Forum Moderator Posts: 32,973 Sapphire Posts by user Forum Profile RuneMetrics Profile

It's definitely worrying where things are headed, especially as they move faster than the controls and regulations etc needed in order to ensure it doesn't hugely negatively impact society.

Overall I think AI is pretty useful in so many different contexts but certainly can'y be left unchecked, especially as it grows in both popularity and intelligence.

I've started using ChatGPT recently (more out of curiosity of its capabilities) during work to save me time on functionality I'm creating for a system we use. Now one angle of that is saying its lazy (as Tuffty said) but I see it more as an additional tool to improve productivity.

Here's an example. About a month ago I was working on some new internal functionality. One portion of that was writing a function to output the duration of a process formatted in a particular way (e.g 1 hour, 3 minutes, 1 second) from a MySQL timestamp stored in the DB.

As I started to write that function I was curious how well ChatGPT would do if I asked it to do it for me, I also thought why would I spend an hour writing a function when AI could write it for me in 20 seconds. As long as I understand completely what code it wrote and how it works, then it's not really a problem. The alternative could be to look on Stackoverflow if someone had done the same thing as there's always something someone has written that gets you either part way to the solution you want or all the way - that's kinda the purpose of libraries, to not have to constantly re-invent the wheel - I wouldn't say that's lazy, it's productive.

The cool thing about ChatGPT (it's the only one I've used) is that it understands context of a conversation. So I tell it what I want, it writes me a function and I can ask it for additional things and it understands. So the first function it gave me didn't take plurals into account (e.g hours vs hour depending if its 1 or more) so after a few exchanges, I got exactly what I wanted.

Joel

Need support? Support Centre | Forum Help

02-May-2023 15:16:53

Feb

2005

Joel

Forum Moderator Posts: 32,973 Sapphire Posts by user Forum Profile RuneMetrics Profile

It's incredibly powerful in that way, saved me an hour of my time so I could be onto the next task delivering something much quicker.

Literally end of day yesterday I and another developer had an issue with something, I asked ChatGPT on how to achieve something as what we did didn't seem to work. It gave me a solution, I ran it and the result was an error. I fed the error back to ChatGPT and it gave me a fix and boom, it was fixed

So it's very useful in at least that context as I have experience in it. That's not to say it's not harmful in others however!

Joel

Need support? Support Centre | Forum Help

02-May-2023 15:19:08

Origin Nexus

Posts: 322 Silver Posts by user Forum Profile RuneMetrics Profile

Averia Light said :
LLMs, like Chat GPT, do have data to work with, but it isn't limited to the responses someone creates and it does have the capability to evolve in real time. Most ChatAI doesn't evolve in real time, with good reason; the internet destroys any ChatAI with a running training model, like Tay and Sydney. I would have thought that Microsoft would have learned their lesson after they allowed Tay public access on Twitter, and within a day, it became a racist asshole. But no, a few years later Microsoft released Sydney onto Twitter, and the exact same thing happened.

Joel said :
As I started to write that function I was curious how well ChatGPT would do if I asked it to do it for me, I also thought why would I spend an hour writing a function when AI could write it for me in 20 seconds. As long as I understand completely what code it wrote and how it works, then it's not really a problem. The alternative could be to look on Stackoverflow if someone had done the same thing as there's always something someone has written that gets you either part way to the solution you want or all the way - that's kinda the purpose of libraries, to not have to constantly re-invent the wheel - I wouldn't say that's lazy, it's productive. I'm curious how it turned out. Because, as you know, it doesn't actually understand what it's outputting, it's only producing code based on the probability of each word being correct in the usage based on it's training model. I've seen people producing working code directly from ChatGPT, but I've also seen ChatGPT output meaningless code that won't even compile. If you're asking for a common function that has been posted on the internet a plethora of time, you're a lot more likely to get an output of fairly decent code. If you're asking for an obscure function that no one has ever produced, you're probably going to have to be very specific in your requests.

02-May-2023 20:31:08

Forums

A.I. Should we be worried?