that's probably why they released stable diffusion, to collect data for stable diffusion inputs which they're going to use to automate prompts and then all the people creating stable diffusion threads on Bot.info every 5 minutes will be out of a job
Not as well paid as the designers and animators who create videos with rudimentary shapes and shading that they will use as sources to diffuse, unfortunately.
It's ridiculous because the prompting is completely useless for the AI, a slot-filling system would be infinitely superior (not only can you do automatic slot-filling for normalfags while having far less likelihood of mistakes, you can also much more easily disable topics you don't like, and even allow manual control over slots to completely bypass prompt understanding errors).
Dialogue systems can generally be classified in 2 types: open and closed. Open means you take any dialogue as input and must answer regardless.
Closed means you only consider a range of concepts, which are managed by slots, and answer based on that.
For example, supposed you have a chatbot for booking flights. This would normally be implemented as a closed dialogue system. The slots considered may be:
- Start date
- End date
- Destination airport
- Origin airport
- Flight class
And so forth.
A dialogue usually proceeds with the user providing basic data such as "I want to go to cancun". The bot then has to understand that cancun should be put in the 'destination airport' slot as 'incomplete data' (because the specific airport is not identified and there may be several). This will be resolved later (possibly through dialogue: the system may ask the followup: would you prefer arriving at airport XXX or airport YYY? to try to fill the slot). To fill more slots, the bot will ask more questions: "when would you like to leave?" and so forth.
When enough data is available, the bot performs a search and returns the top-N results. If the user selects one, the flight can proceed to booking. If not, the bot attempts to fill more slots, or to correct slots, based on user input.
>believing that the "input part" is the most important thing and that the AI creators won't use your inputs to optimize the input part of the AI so that it gives better results with dumber inputs
Yeah, ngmi.
Coomers always pave the way in software revolutions and the furries are winning the AI game followed by tranime fag's then at dead last normal coomer fag's who try and make real women.
>job
go back worktard
>check Bot.info
>frog frog frog /misc/ frog...
Same as always, huh.
AI poweruser
we can simply have an AI for coming up with prompts for other AIs
that's probably why they released stable diffusion, to collect data for stable diffusion inputs which they're going to use to automate prompts and then all the people creating stable diffusion threads on Bot.info every 5 minutes will be out of a job
>will be out of a job
How to lose something you never had?
can never lose a job if your job description is NEET
Not as well paid as the designers and animators who create videos with rudimentary shapes and shading that they will use as sources to diffuse, unfortunately.
> I'm gonna proooooooompt
It'll be like chess. We'll have tournaments where the two best prompters in the world go head to head.
Nah, idea guys are still worthless
It's ridiculous because the prompting is completely useless for the AI, a slot-filling system would be infinitely superior (not only can you do automatic slot-filling for normalfags while having far less likelihood of mistakes, you can also much more easily disable topics you don't like, and even allow manual control over slots to completely bypass prompt understanding errors).
slots? what is this, las vegas?
Dialogue systems can generally be classified in 2 types: open and closed. Open means you take any dialogue as input and must answer regardless.
Closed means you only consider a range of concepts, which are managed by slots, and answer based on that.
For example, supposed you have a chatbot for booking flights. This would normally be implemented as a closed dialogue system. The slots considered may be:
- Start date
- End date
- Destination airport
- Origin airport
- Flight class
And so forth.
A dialogue usually proceeds with the user providing basic data such as "I want to go to cancun". The bot then has to understand that cancun should be put in the 'destination airport' slot as 'incomplete data' (because the specific airport is not identified and there may be several). This will be resolved later (possibly through dialogue: the system may ask the followup: would you prefer arriving at airport XXX or airport YYY? to try to fill the slot). To fill more slots, the bot will ask more questions: "when would you like to leave?" and so forth.
When enough data is available, the bot performs a search and returns the top-N results. If the user selects one, the flight can proceed to booking. If not, the bot attempts to fill more slots, or to correct slots, based on user input.
Do you think that looks pretty? There are shitty blender renders out there that have more soul that this garbage.
>believing that the "input part" is the most important thing and that the AI creators won't use your inputs to optimize the input part of the AI so that it gives better results with dumber inputs
Yeah, ngmi.
this shit will be as stagnant as VR in a few more months.
More coom potential with this than VR headset's.
Coomers always pave the way in software revolutions and the furries are winning the AI game followed by tranime fag's then at dead last normal coomer fag's who try and make real women.