Noticed everyone keeps calling open source AI 'not real AI' at my local tech meetup in Seattle

I was at a small tech meetup in Seattle last week and this guy kept saying open source models like Llama 3 are just toy projects compared to GPT-4. He said they can't handle real world tasks, which I thought was weird since I've been running a local Llama 3 70B model for my dental office scheduling for two months now. It handles appointment reminders, insurance code lookups, and even generates follow up emails without me paying per token. I get that OpenAI's stuff is polished but calling open source 'not real AI' just dismisses how fast things have moved since Meta released it in April. Has anyone else run into this gatekeeping attitude at industry events?

2 comments

2 Comments

piper_dixon451mo ago

That tech bro mentality is exhausting, glad you called it out.

skyler_anderson651mo ago

Honestly I read a breakdown from some ML researcher recently who tested Llama 3 70B against GPT-4 on actual business workflows, and it scored within 10% on most tasks. People forget that for specific use cases like scheduling or code lookups, open models can be way more efficient since you control the context and fine tuning. Ngl the "real AI" gatekeeping feels like classic tech bro one-upsmanship, especially when the cost difference is literally thousands of dollars a month for small businesses.