OpenAI, Anthropic, and Meta | Analyzing the AI Frontier with Zvi

Watch Episode Here

Video Description

This isn't news, it's analysis! Nathan Labenz sits down for an with Zvi Mowshowitz, the writer behind Don't Worry About the Vase to talk about the major players in AI over the last few months. In this extended conversation, Nathan and Zvi debate if AI has attained the intelligence of a well-read college graduate (per OpenAI's Jan Leike), a live player analysis, and the role of independent red teaming organizations. If you're looking for an ERP platform, check out our sponsor, NetSuite: http://netsuite.com/cognitive

Definitely also take a moment to subscribe to Zvi's blog Don't Worry About the Vase
(https://thezvi.wordpress.com/) - Zvi is an information hyperprocessor who synthesizes vast amounts of new and ever-evolving information into extremely clear summaries that help educated people keep up with the latest news.


TIMESTAMPS:
(00:00) Episode preview
(03:15) Is AI as intelligent as a college grad?
(07:45) Memories and context processing
(15:45) Sponsor: NetSuite | Omneky
(17:13) Is AI as intelligent as a college grad? cont'd
(20:47) Strengths and weaknesses of AI vs human
(31:05) OpenAI Superalignment
(37:23) The relationship between OpenAI and Anthropic
(44:31) Anthropic's security recommendations and adversarial attacks
(50:50) Is OpenAI using a constitutional AI approach?
(01:01:26) Context and stochastic parrots
(01:10) Is more context better?
(01:15:29) Should Nathan work at Anthropic?
(01:21:35) Google DeepMind's RT-2
(01:27:47) Multi-modal Med-PaLM
(01:31:50) Speculating about Gato
(01:35:10) Skepticism about Med-PaLM usage in radiology
(01:41:37) Llama 2 - what is going on at Meta??
(01:51:14) Llama 2 vs other models
(01:55:29) Who are the live players?
(02:01:38) China's AI developments
(02:02:41) Character AI and inflection
(02:05:26) Replit as the perfect substrate for AGI
(02:10) AI girlfriends
(02:18:53) AI safety: The White House
(02:25:43) Bottlenecks to progress
(02:35:27) Can new players influence AI policy?
(02:39:00) Liabilities
(02:47:54) Independent red teaming organizations
(02:57:18) Mechanistic interpretability

LINKS
- Zvi's blog: https://thezvi.wordpress.com/
- Google's RT-2: https://www.deepmind.com/blog/rt-2-new-model-translates-vision-and-language-into-action
- https://www.deepmind.com/publications/a-generalist-agent
- Llama 2: https://ai.meta.com/llama/
- Adversarial attacks paper: https://arxiv.org/abs/2307.15043

RECOMMENDED THE COGNITIVE REVOLUTION EPISODES:
- AI Safety Debates with Zvi https://www.youtube.com/watch?v=5yM7fIfxYV8
- Mechanistic Interpretability with Arthur Conmy https://www.youtube.com/watch?v=Y5Pzch7_8MQ&t=5670s
- Med-PaLM with Vivek Natarajan https://www.youtube.com/watch?v=nPBd7i5tnEE

X:
@labenz (Nathan)
@thezvi (Zvi)
@eriktorenberg (Erik)
@cogrev_podcast

SPONSORS: NetSuite | Omneky

-NetSuite provides financial software for all your business needs. More than thirty-six thousand companies have already upgraded to NetSuite, gaining visibility and control over their financials, inventory, HR, eCommerce, and more. If you’re looking for an ERP platform: NetSuite (http://netsuite.com/cognitive) and defer payments of a FULL NetSuite implementation for six months.

-Omneky is an omnichannel creative generation platform that lets you launch hundreds of thousands of ad iterations that *actually work* customized across all platforms, with a click of a button. Omneky combines generative AI and real-time advertising data. Mention "Cog Rev" for 10% off.


MUSIC CREDIT:
MusicLM