Claude 3.7 is More Significant than its Name Implies (ft DeepSeek R2 + GPT 4.5 coming soon)
Claude 3.7 is here, hot on the heels of Grok 3 and a host of other developments, but how good is it really? And what does it say about the next few months in AI? I’ve read the papers, played with the model for hours, and benched it on Simple. Things aren’t slowing down. Plus the latest in humanoid robots, led by Helix and freaked out by Protoclone. And reports of GPT 4.5 and DeepSeek R2.
GraySwan Competition!
AI Insiders ($9!):
Chapters:
00:00 – Introduction
01:25 – Claude 3.7 New Stats/Demos
05:22 – 128k Output
06:13 – Pokemon
06:58 – Just a tool?
09:54 – DeepSeek R2
10:20 – Claude 3.7 System Card/Paper Highlights
17:18 – Simple Record Score/Competition
20:37 – Grok 3 + Redteaming prizes
22:26 – Google Co-scientist
24:02 – Humanoid Robot Developments
3.7 Release Notes:
vs o3 and Grok 3:
Extended Thinking:
System Prompt:
System Card:
Unfaithful CoT:
Original Constitution:
Responsible Scaling Policy:
Amodei and Hassabis:
400 Weekly Users:
Grok 3 Jailbroken:
Google Co-Scientist:
But Hassabis Says Years Away:
DeepSeek R2 Reuters:
Protoclone:
Helix:
TechTrance:
GPT 4.5 Soon:
Altman roadmap:
Non-hype Newsletter:
Podcast:
Interesting
nearly half an hour? you’re spoiling us philip
Woo another video! Thanks!
I think any concern I’ve had of a *plateau* is now gone.
Yesssss!!!! omggg was waiting for this!!! hoep you are feeling better,! Just made my week!
Sam said GPT-5 is smarter than him in a press conference. But GPT-5 is going to be released in May supposedly. Then he also made claims of AGI in 2025. With 4.5 being an AGI moment. I am not sure how to feel.
This is never not the first thing in my YouTube recommended when it comes out
YES! Best AI content on YouTube
Video on grok 3 pls. You are the only person I trust to give me accurate information
I’m the same, I was expecting the new video to be about Grok 3.
sadly he lives in the uk and most likely reads bbc, the guardian and the such and really believe them to be neutral. he said something last video that raised a red flag sadly. these people live in 1984 and they don’t know. the frog in boiling water. he didn’t talk yet about grok because he don’t like elon
Can you share your code for the convorecy segmentation? I’m super curious. 🙂
was constantly refreshing channel video page.
Was waiting for this.. Let’s go🔥
I wonder how the newest quantum computing breakthroughs from Microsoft will affect the AI scene, it seemed pretty major from what I’ve heard
I’m convinced he’s got insider knowledge with the timing on these past videos
When I saw the video from anthropic in my feed the second thing that came on my mind (after “oh wow, a new model?”) was “WE GETTING A NEW AI EXPLAINED VIDEO 🗣️ 🔥🔥”
Was missing your videos, Philip. Good to see you are well.
Nice
So glad you’re here creating videos—refreshingly different from the overhyped finance gurus or the ‘just trust me, I know what I’m talking about’ types.
You had me worried. No Grok 3 video? Glad you’re alive 🙂
14:38 Incidentally, the model did get the episode list *almost* right (Episode 7’s actual title is “The Lowlands”, not “The Lowering”). So, unlike the paper’s claim, it did not in fact remember them all correctly, but it also definitely didn’t just hallucinate the list. It’s only the CoT where it was underestimating its own knowledge.