Discover more from Exponential View by Azeem Azhar
📺 I caught up with Sam Altman...
We speak about Sam's mentors, motivation to do his job, the next paradigm in AI, and the safety challenges in building ever more powerful models
Hi, it’s Azeem.
Three short months ago, I got the opportunity to join Sam Altman on stage at University College London during his World Tour. In the course of thirty days, Sam visited 20 countries across five continents. After London, where I met him, he was flying further east, across Europe to India — his focus and dedication that of a man on a mission of his lifetime. He was strikingly open, vulnerable, and eager to listen — even to those with drastically opposing views.
You can now watch my conversation with Sam below…
Spread the word:
Watch the video and share it with your network,
Leave a comment on the video — tell us what you took away from the chat.
I first met Sam in 2019, right after he introduced GPT-3 as a “baby step” on the development curve of AI. In our conversation from that year, we were still talking about the abilities of AI as future potential. It wasn’t until GPT-3.5 and then GPT-4 were made available that we all got the first-hand experience of what AI actually was capable of — and how disruptive it would be.
Our conversation from earlier this year, left me thinking about the manner in which AI leaders — Sam and his peers — have taken it upon themselves to lobby for safe regulatory frameworks. While I don’t believe it’s up to technology leaders to steer AI towards the best outcomes, this is an unprecedented approach in the tech industry.
It is indisputable that they feel a great deal of pressure and responsibility to act as gatekeepers of AI’s progress. In our conversation, Sam shared that his team spent eight months between the creation and launch of GPT-3.5 due to safety concerns - the longest they ever spent on a pre-release.
We also discussed…
Sam’s inner drive to work at OpenAI,
Figuring out the ‘next paradigm’ that can help us unlock new science and productivity gains,
The alignment and safety challenges in developing advanced models.