medians wrote:

Okay, I'm even more worried about this Chatbot thing now.

cs3868895 wrote:

It may not be in Tolors but it seems that allowing bots to control may actually be rejected a long time ago. I haven't gotten confirmation just yet, this is just speculation.

Yeah, it might just spit out random stuff or inaccurate stuff now.

Knightbot63 wrote:

Chatbots like ChatGPT may not be accurate. Unlike human contacting, it's more unstable and inaccurate.

Knightbot63 wrote:

Chatbots like ChatGPT may not be accurate. Unlike human contacting, it's more stable and accurate.

cs3868895 wrote:

(#23)

Knightbot63 wrote:

Chatbots like ChatGPT may not be accurate. Unlike human contacting, it's more stable and accurate.

more unstable and inaccurate you mean?

Basilikos wrote:

I know ChatGPT-powered things very cool elon musk moment but please, do not put it everywhere. A human is a billion times better at understanding you than stupid automatically generated letters on the screen.
Why do we even need this?
The bot is trained for all kinds of tasks such as changing usernames, moderating and others. Why do you think that it won't get confused one time and accidentally launch several thermonuclear warheads to French Polynesia? A human can distinguish better than a bot, at the moment.
Sorry if those silly essays about the colour of apples mislead you; but humans win this round.

cs3868895 wrote:

A bot cannot understand another bot's work too not to mention.

Basilikos wrote:

cs3868895 wrote:

A bot cannot understand another bot's work too not to mention.

Could you imagine all those soyboty capitalist elon musk suckers using Chat-GPT to write stuff like

“Hello dear SussyTeam
please change my username for it makes me spill vomit
best regards, totally human person”

And then the bot's just like

“Hello user
I am not trained for your speech so I didn't understand you
sorie”

cs3868895 wrote:

snip

Basilikos wrote:

cs3868895 wrote:

snip

Filtering data you mean?
Yeah…

Basilikos wrote:

Ok so turns out that I'm not a machine learning guy so I can't really talk about this in depth

cs3868895 wrote:

Computer fish brain:

In order to make it really, really precise, you gotta play with the chatgpt first. Making it even more precise, you need to put data, the code is the data. What are the possible answers to the questions? What is right? What belongs here?
For every possible answer and how precise, it depends on how much data the person puts, and the person themselves. There's gotta be bias, and if there is a mistake the ai will repeat.



Given how small the scratch team is especially at engineers, I don't think this will be a high priority, and even if it is, time consuming.

Making ai's is not easy, and keep in mind the current one is for paid subscribers. That can waste alot of money. It's also, a bot. To the public, and you can't escape bias.

Mryellowdoggy wrote:

What would the bot even be set up to do? I think that the AI stuff is cool, but it definitely isn’t at the level to replace customer service yet, unless it was only for very basic tasks.

cs3868895 wrote:

Mryellowdoggy wrote:

What would the bot even be set up to do? I think that the AI stuff is cool, but it definitely isn’t at the level to replace customer service yet, unless it was only for very basic tasks.

what does customer service have to do about this topic

Mryellowdoggy wrote:

cs3868895 wrote:

Mryellowdoggy wrote:

What would the bot even be set up to do? I think that the AI stuff is cool, but it definitely isn’t at the level to replace customer service yet, unless it was only for very basic tasks.

what does customer service have to do about this topic

Sorry if I read the topic wrong, but Contact Us is basically for the Scratch Team to help Scratchers, wouldn’t it?