News on China's scientific and technological development.

Overbom

Brigadier
Registered Member
Indeed, but what I would say to you is that it’s they quality if the human that matters now. Without humans it’s just AI training AI and ultimately those AI’s were based on human annotations all the way back.

Nowadays it’s the experts that add the key annotations and train the lower cost annotators, so in the end it’s all down to your education system.
A relatively few smart people are needed at the beginning and then the AI takes over, with some human input here and there. Even data annotation itself, which is the most labour intensive process, increasingly becomes automated nowadays.

Just to give you an idea, OpenAI has just 375 employees. If you consider their different products and administration costs, we could generously assume a maximum of 100 people are involved in their language model projects.

Of course they hired external contractors, but I wouldn't expect that number to be something crazy like 10000 or something like that. They would need their core data annotation team (20 people?) to supervise them in order to ensure a consistently high quality data annotation dataset. In general, human data annotation or supervised learning quickly finds itself in big problems when the time to scale comes
 

Andy1974

Senior Member
Registered Member
A relatively few smart people are needed at the beginning and then the AI takes over, with some human input here and there. Even data annotation itself, which is the most labour intensive process, increasingly becomes automated nowadays.

Just to give you an idea, OpenAI has just 375 employees. If you consider their different products and administration costs, we could generously assume a maximum of 100 people are involved in their language model projects.

Of course they hired external contractors, but I wouldn't expect that number to be something crazy like 10000 or something like that. They would need their core data annotation team (20 people?) to supervise them in order to ensure a consistently high quality data annotation dataset. In general, human data annotation or supervised learning quickly finds itself in big problems when the time to scale comes
Well my company had 4 managers, running a team of 250 annotators, I assure you it is scalable.
 

Overbom

Brigadier
Registered Member
Well my company had 4 managers, running a team of 250 annotators, I assure you it is scalable.
250 annotators I assume in poor countries? Kinda impressive if they could handle all that output and still maintain high quality datasets. From what I have heard, there are big problems when so few people try to manage the output of so many people while keeping good quality

Have heard some horror stories about Indian contractors lol.
 

Andy1974

Senior Member
Registered Member
250 annotators I assume in poor countries? Kinda impressive if they could handle all that output and still maintain high quality datasets. From what I have heard, there are big problems when so few people try to manage the output of so many people while keeping good quality

Have heard some horror stories about Indian contractors lol.
Serbia and PH. I did start in India, but left very quickly.

Here is the workflow…

1. Take picture on mobile app
2. Image QA on platform
3. Tag or Mask objects
4. Annotation QA (image level)
5. Annotation QA (dataset level, my invention)

4 and 5 are done by experts of different quality.
Teams are split up with annotators and QA members in each, with a ration of 3:1. Managers hire and train both.
 

CMP

Senior Member
Registered Member
250 annotators I assume in poor countries? Kinda impressive if they could handle all that output and still maintain high quality datasets. From what I have heard, there are big problems when so few people try to manage the output of so many people while keeping good quality

Have heard some horror stories about Indian contractors lol.
Outsourcing anything to India is an absolute nightmare. I speak from personal experience in STEM.
 

Overbom

Brigadier
Registered Member
Serbia and PH. I did start in India, but left very quickly.

Here is the workflow…

1. Take picture on mobile app
2. Image QA on platform
3. Tag or Mask objects
4. Annotation QA (image level)
5. Annotation QA (dataset level, my invention)

4 and 5 are done by experts of different quality.
Teams are split up with annotators and QA members in each, with a ration of 3:1. Managers hire and train both.
Ok I see now, this workflow seems reasonable on how to handle the workload in an efficient way by allocating different tasks to different teams each specialised on their own field.

Btw, I don't know if you have seen it yet but the Alpaca paper seems revolutionary on its implications. This might be a way for China to quickly catch-up to the frontrunners easily. According to that paper from Stanford, they were able to quickly and cheaply fine-tune their LLaMA model with text-davinci-003 lol

OpenAI's perceived advantage doesn't seem to be that much with this new revelation. It means that any service that provides wide access to its AI is potentially also fine-tuning other people's models.

Remember all these millions (billions?) that OpenAI had to spend to train its model? Yeah, about that...
 

Bellum_Romanum

Brigadier
Registered Member
And then there are retarded articles such as this

Please, Log in or Register to view URLs content!
If the pre-requisite for inventing such "inventive" ideas and leading tech of tomorrow is once again based on this canard of "democracy, freedom" b.s. then why couldn't other countries, especially in Europe could ever come up with ingenious tech like ChatGPT? How about India? the so-called largest democracy in the world, with size and population that has now exceeded China, yet, I don't know when was the last time that country has managed to come up with INNOVATION that caught the world by storm as some other apps that has come from SEE SEE PEE infested China.

The propaganda from America is f...ng annoying, hollow, and utterly contemptible. It treats its readers as dumb mother f...kers and gullible sheep unable to discern through the clear b.s. this sort of rag that pride itself as some sort of beacon of great journalism.. What an utter crock of s..t.
 
Top