I want to learn robotics, but my friends have suggested I learn coding first before pursuing a career. So, I am enrolling myself at Moonpreneur to build a strong foundation.
While the growth metrics are impressive at first glance, one must consider the broader context of Fireworks AI's business model. Operating essentially as an optimization layer for open-source models, they face significant challenges: dependency on external model development, fierce competition from both tech giants and nimble startups, substantial infrastructure costs, and limited differentiation opportunities. The touted "100x growth" in 6 months, while noteworthy, may reflect more about the timing of their launch during 2023's AI boom than sustainable competitive advantage. In an industry where margins are constantly under pressure and switching costs are low, the real test will be maintaining growth while building lasting technological moats beyond mere infrastructure optimization. The GPU-intensive nature of their business model combined with reliance on others' intellectual property presents interesting challenges for long-term sustainability.
So looks like what they do is Llama hosting? Can't you do the same with aws and fire up a llama instance? Or run it on Groq cloud and get much faster inference? I might be missing something here.
Instead of you hosting your own models and applications on servers, they manage everything and provide a platform to run these specialized AI models. Think of it as AWS, but just for AI.
You miss interpreted what her meant, everyone knows PyTorch is deep learning framework. Models that trained in PyTorch cannot directly transfer to other frameworks. You have to target specific framework to work on those models hence PyTorch models.
For sure, but doesn't mean that AI isn't a thing. Just a way to leverage the huge amounts of human-produced data that the internet has produced -- both understand/analyze it and produce similar data synthetically. Humans tend to overreact to new, exciting technologies though
@@MathewKalarikkal-t7y you definitely could, but the time you will spend learning how to build with no code vs just buying a cheap ready made mvp 🤷🏻♂️ guess its an opportunity cost thing
I second this none of these ai companies solve a real problem Amazon all the tech before solve real world problems stop telling us about search and data mining google an all the others before doing this come up with some real company. These people are so flipping lazy
These guys do solve real problems. Just not for the average Joe, but for companies like Bank of America, if they're trying to use Generative AI in their customer service to. Building an AI team from scratch at each company that wants to use it is expensive and time consuming.
This is 100% fraud. This is the first of many fad startups that you’ll be seeing. The best possible case for “companies” like this is to take the money and run before these actual companies realize they can do all of this with 1 or 2 people. Saying “you need 100 people” to do anything is fucking wild. I could take 3 of my best guys and take over a country. Give me a couple weeks and I’ll build you an agent that can do everything this startup can do without a 500 million dollar eval
Wonderful to see her interview. I saw how the website changed and evolved
I want to learn robotics, but my friends have suggested I learn coding first before pursuing a career. So, I am enrolling myself at Moonpreneur to build a strong foundation.
Sup everyone,
always remember that Healthy body is the greatest treasure . May god keep you healthy always ❤
Good Luck Sister ❤🎉🎉🎉
awesome story ~
Awesome interview & founder. Love how fireworks integrates throughout with many open source frameworks.
While the growth metrics are impressive at first glance, one must consider the broader context of Fireworks AI's business model. Operating essentially as an optimization layer for open-source models, they face significant challenges: dependency on external model development, fierce competition from both tech giants and nimble startups, substantial infrastructure costs, and limited differentiation opportunities. The touted "100x growth" in 6 months, while noteworthy, may reflect more about the timing of their launch during 2023's AI boom than sustainable competitive advantage. In an industry where margins are constantly under pressure and switching costs are low, the real test will be maintaining growth while building lasting technological moats beyond mere infrastructure optimization. The GPU-intensive nature of their business model combined with reliance on others' intellectual property presents interesting challenges for long-term sustainability.
So looks like what they do is Llama hosting? Can't you do the same with aws and fire up a llama instance?
Or run it on Groq cloud and get much faster inference? I might be missing something here.
Still no idea what they do after 12 mins listening to her talk
Instead of you hosting your own models and applications on servers, they manage everything and provide a platform to run these specialized AI models. Think of it as AWS, but just for AI.
wonderful from one of the creators of pytorch
PyTorch is not a model, it’s a python-based deeplearning framework for building AI models.
You miss interpreted what her meant, everyone knows PyTorch is deep learning framework. Models that trained in PyTorch cannot directly transfer to other frameworks. You have to target specific framework to work on those models hence PyTorch models.
Fast and Furious growth.
Often these fail as fast as they ascended
I’m still confused so they’re like a third party that manages Ai for other companies so they don’t have to have an in house team ?
Iconic!!
great interview
Bubble
I said it first
maybe
For sure, but doesn't mean that AI isn't a thing. Just a way to leverage the huge amounts of human-produced data that the internet has produced -- both understand/analyze it and produce similar data synthetically. Humans tend to overreact to new, exciting technologies though
Do you think I can build a unicorn startup using no code tools?
Unlikely but possible. What are you using, bubble?
You could, but i’d recommend just buying a ready made software to test the market rather than building one from scratch.
Nah bruh, you could build a POC or MVP with nocode. But to scale you would need to build it
@@MathewKalarikkal-t7y you definitely could, but the time you will spend learning how to build with no code vs just buying a cheap ready made mvp 🤷🏻♂️ guess its an opportunity cost thing
Unicorn founders don't deliberately build company with a unicorn goal.
Wow
Move fast(.)
I think its a fraud.
I second this none of these ai companies solve a real problem Amazon all the tech before solve real world problems stop telling us about search and data mining google an all the others before doing this come up with some real company. These people are so flipping lazy
These guys do solve real problems. Just not for the average Joe, but for companies like Bank of America, if they're trying to use Generative AI in their customer service to. Building an AI team from scratch at each company that wants to use it is expensive and time consuming.
This is 100% fraud. This is the first of many fad startups that you’ll be seeing. The best possible case for “companies” like this is to take the money and run before these actual companies realize they can do all of this with 1 or 2 people. Saying “you need 100 people” to do anything is fucking wild. I could take 3 of my best guys and take over a country. Give me a couple weeks and I’ll build you an agent that can do everything this startup can do without a 500 million dollar eval
I’m a user, good service imo
Who cares what you think if you’ve never tried it and don’t have experience in the industry?
Fradu no 1.
I can start another AI Startup exactly like your model and installed in Nvidia Inference Network Microservices. All by AI Agents.
what are you waiting for?
ok. and revenue?
She looks a little like Jensen Huang…😂