NVIDIA said they built the dual H100 with the NV link for that. And that's 160 gigabytes of VRAM, which would imply a 200 gigabyte model. That's a 100 gigabyte model, 200 billion. Something that can pass all these exams. They're designed to be reasoning machines, not fact machines. So hallucination is in a hallucination. It's just if you don't know something sometimes, you might just make it up or do a post-talk rationalization. Like Cicero by Meta was an amazing paper, they took eight language models and got them to interact with each other and beat humans at the game of diplomacy. These are the things we have to
Emad Mostaque is the Co-Founder and CEO @ StabilityAI, the parent company of Stable Diffusion. Stability are building the foundation to activate humanity's potential. To date, Emad has raised over $110M with Stability with the latest round reportedly pricing the company at $4BN. Investors include Coatue, Lightspeed, Sound Ventures, OSS Capital and Airstreet Capital, to name a few. Prior to Stability, Emad was in the world of hedge funds, that was until his son was diagnosed with autism and he left to make a difference in the space and help find treatments and solutions.
In Today's Episode with Emad Mostaque We Discuss:
1.) From Hedge Funds to Finding Treatments for Autism to Leading the World of AI:
- How Emad made his way from the world of hedge funds to founding one of the leading AI companies of our time?
- How did Emad find a solution to parts of his son's autism with a $6 drug?
- How does Emad believe we can use AI to solve the majority of medical problems today?
- What does the future of healthcare look like with AI at the centre?
2.) Models: What is Real? What is False?
- Why no models today will be used in a year?
- Why all models are biased and how do we solve for it?
- Why hallucinations are a feature and not a bug?
- Why the size of your model does not matter anymore?
- Why will there be national models specified to cultures and nations? How is this implemented?
3.) Who Wins: Startups or Incumbents:
- Why does Emad believe there will only be 5 really important AI companies? Which will they be?
- How does Emad review Google's AI strategy following their news last week? Was their integration of Google and Deepmind recently a success?
- How does Emad assess Meta's AI strategy? Why does Zuckerberg now acknowledge the metaverse play was a mistake?
- How does Emad evaluate the approach taken by Amazon? Why are they the dark horse in the race?
- What can startups do to get a meaningful edge on the large incumbents? How do they compete with their distribution?
4.) The Next 12 Months: What Happens:
- Why does Emad believe the .ai bubble will be bigger than the dot com bubble?
- Why does Emad believe that the biggest companies built-in AI in the next 12 months will be services-based companies? How does the ecosystem look if this is the case?
- Why will India and emerging markets embrace AI faster than anyone else? What happens to economies that have large segments reliant on freelance work that AI replaces?
- Why will we see the death of many large content publishers and media companies? What does Emad mean when he says we will see the rise of "AI first publishers"?
5.) Open or Closed: What Wins:
- Why does Emad believe we must be open by default? Why does open win?
- Why does Emad side with Elon and believe we must pause the development of AI for 6 months?
- How does Emad evaluate the leaked memo from Google stating that neither Google nor OpenAI are ahead? What does this mean for the AI ecosystem?
- Where will the best AI talent concentrate? What do companies need to do to win the best talent?
