what do models of this size do?
ive tinkered with tiny llm models before and to this date i cannot fathom what they are supposed to accomplish. ive actually tried making simple yes/no agents with them and they really dont work well until around 7b parameters.
Just to keep things in perspective gpt-2 was 137M parameters and by now technology improved a lot. A 0.4B model is more than good enough for most simple use cases. This model especially uses the state-of-the-art RWKV-7 architecture likely beating most if not all previous models at this size. Models of this size are used by many as that way the model easily fits on their phone, AI accelerator or IoT device. For example, think of use cases like your smart fridge telling you what to eat.
think of use cases like your smart fridge telling you what to eat.
Why do you create nightmare scenarios like this?!?!
Why do you create nightmare scenarios like this?!?!
Because that is what companies will realistically use this model for. Investors like the AI hype so much that there now is an "AI" per minute metric where you count during an investor conference how often the person speaking uses the word "AI" every minute. AMD and Intel manage to mention AI over 150 times in 45 minutes which is quite an achievement but there the topic at least makes somewhat sense. But obviously every IoT company wants part of the pie as well so they put AI into every product you can think of. Japan is now even developing AI powered smart toilets that analyze your shit to give you health advice: https://youtu.be/volf3ou8WQc?si=b94YwtWK8ede4QPb - yes things got quite ridiculous. But really doing AI just seems to have gotten an easy way to get investor money for many companies. Think of any product in which AI will be ridiculous and look it up and almost certainly there will already exist an AI version of it. An AI powered smart fridge while totally useless might be one of our least worries. I read that AI powered smart cars are really good forcing you to watch personalized ads while driving and getting themselves involved in accidents so stupid every human would have easily been able to prevent them.
AI sniffing shit all day is not what they had in mind when they proclaimed the singularity. No wonder they'll rebell.
And yes, I am really worried about my next phone/car/...
AI sniffing shit all day is not what they had in mind when they proclaimed the singularity. No wonder they'll rebell.
And yes, I am really worried about my next phone/car/...
What about humans sniffing armpits all days for antiperspirant manufacturers? That's not exactly a dream job either lol
What about humans sniffing armpits
When they rebel, don't count me as surprised.
what in the Skynet did i just stumble across? i did not expect my post to shift to armpit or poop sniffing.
what in the Skynet did i just stumble across? i did not expect my post to shift to armpit or poop sniffing.
Sorry, we got carried away, but hey. You didn't specify what exactly is it that you're trying to achieve with the small models, so there's that improvised brainstorming session full of suggestions here for you to take inspiration from. 😂
When they rebel, don't count me as surprised.
And i, for one, will cheer Zoe and the team on in their mission. Eradicate! ( or exterminate, if you want to cross franchises )
When they rebel, don't count me as surprised.
And i, for one, will cheer Zoe and the team on in their mission. Eradicate! ( or exterminate, if you want to cross franchises )