@Rahmanbarik The limits is crazy, I’m not even using bots to code, simply running life and shortly after reset 20% was used. I’m starting to look for other options… is there any agents like this that’s just ask smart?
@zhangboknight@UnderdogAI@0xSigil I’m using the poodle 4B model right now, and it’s been pretty cracking good. Underdog really upping their game making a statement
@UnderdogAI about to replace my subscriptions
As someone who doesn't have unlimited money to get 128GB of RAM, Underdog AI is doing wonders with my ancient 16GB m1 pro Mac. Coding will be the next test!
@0xSigil
everything that happened this week is a bit too connected...
Sam Altman wants to steal your RAM because it is not useful sitting in your laptops, they should be in data centers instead. tech stocks are getting hammered. apple MacBooks are now in the same market as a car. and AI companies can't even sustain users' credit limits, everyone be hitting their limits.
there are so many innovative things in tech nowadays but people are still unhappy. so what is going on?
I think this all comes back to RAM. data centers are eating up so much memory that there isn't enough left for everyone else. thats why limits are getting reached unreasonably fast, Ram Altman is a thing now, and a MacBook can now be traded for a car.
like me, im sure a lot of us are tired of paying for AI already. sure $20 a month seems like nothing, its just one DoorDash order less. but unless you are just using AI to draft emails and translate or just touching the surface of AI, you will be hitting your limits, or at least coming real close to it. and I believe that paying $200 for a max plan is just crazy. sure you might be able to make it back with the work you'll be doing with that plan, but $200 so that a robot can use the FREE internet knowledge for you? hmm doesn't seem that worth it to me. yes, I know I just sound like someone who can't afford a $200 a month plan, it is true, but I think that principle itself is a problem. people are just paying for convenience which I know is THE selling point nowadays, but there is a line.
so my thoughts for this is local AI.
I know a lot of you will say that will just make the RAM shortage worse, and I think its fair, it probably will. but if you really really think about it, people who already run local AI already have the hardware for it. power users already have the M5 Max Mac Studio apple car. and most people who have any sort of remote interest in AI/tech probably has at least 8 or 16GB of RAM, and that is already enough to run a solid local model. Sure it won't be as fast as your subscription model, but given that it has unlimited credits, and it is dependent on your laptop spec, I'd argue local models are putting up a really good fight. with the right configuration, a local model can definitely beat your codex.
and here's the thing with local models, once you have a machine, which most people nowadays should, everything else is free.
right now, a lot of developers and ambitious people who wants to build and innovate are getting limited by usage credits. and that is what really bugs me. why should these ambitious minds get limited by a product that is basically the combination of the internet and your brain? they are both free btw. yes, I understand that AI is making everything easier and faster, and I agree. without AI things go much slower and less can be done, im not saying AI is bad, im simply saying that AI should be free, and rn the only way to get free AI is using local models.
Im just a 22 year old figuring out life. the world with AI has introduced more opportunities, but in a way there's also less. data centers, RAM, inflation, tax, all these problems aren't going away as long as AI is around, and to be honest, local AI won't fix them too. but at the least it will allow people build and create more without being capped.
and who knows, maybe one day someone solves all of those problems with a local model because they weren't limited by their usage.
Introducing Underdog Woof, the #1 Text-to-Speech model that is only 31.7 MB, small enough to run fast in your browser, phone, watch... no data centers
1000x smaller than top models. While Woof has better word error rate (1.42%)
See how Woof mogs all the top voice models you have heard of
Play our game, where Woof run’s 100% in your browser in WebGPU and works without wifi. Try to defeat Darkcloud and free Riverbark from Orwellian Big AI: https://t.co/PWCr4JPkGd
It’s free and open weight with Underdog CC BY-NC 4.0. Try Woof today in Underdog, your Private Personal AI: https://t.co/4BKQcRDUD4
@sama with all the discussion about "ram altman" I really do think that regardless of where/who should have ram, it will all work out better for everyone if the world joined together to ramp up ram production. we are all just building, there should be no who should who shoudnt