Misha Laskin, co-founder and CEO of ReflectionAI, discusses the launch of Beam, a 500 billion parameter open-weight reasoning model. The conversation explores the challenges and strategies behind developing open AI models and their potential impact on the industry.
Key points
Misha Laskin explains the mission of ReflectionAI to build advanced open intelligence and make it widely available. 1:12
The challenges of scaling AI models include computing resources, talent acquisition, and data volume. 2:05
ReflectionAI's strategic shift to building their own open models was driven by the lack of high-quality Western open models. 6:58
Beam's efficiency is attributed to a strong pre-training base and extensive reinforcement learning. 20:00
Open models are expected to capture the majority of global token demand. 25:04
The competitive landscape of AI models is becoming increasingly intense. 31:52
The importance of open models in cybersecurity and safety is highlighted. 45:46
The potential of AI to accelerate scientific progress is significant. 56:26
Quotes
With enough eyes, all bugs are shallow.
The default state of the world is actually that openness is safety.
Open models are like a Trojan horse for the infrastructure they bring with them.