- AI For Humans - The Newsletter
- Posts
- Jean Claude Vanthropic 3 | Daily Drip
TGI-Tuesday, am I right internet friends?!
Monday was so slow, we had to send it upstate and have it taken behind a barn, and sent to eternally play with all the other Mondays that ever loved it growing up…
Sorry. That’s, terrible. Way too harsh an intro for your inboxes. Let’s consult the oracle for a better opener:
Witness my desperate attempt to amortize my month of Google Gemini Ultra Pro
Gemini suggested: Monday was so slow, the clocks were going backwards!
Beyond the “eesh”, it’s a flawed premise, because Monday saw the release of THREE NEW Claude 3 models… so maybe it’s my fault, not Google’s?
Quick, Gemini, let’s HYPE yesterday uuuUUUP!
Aaaaaand Gemini suggested, wait for it —
Claude 3 Opus Monday Madness: A Day That Rewrote Time
I don’t hate it. I just hate myself for paying for it.
Well throw another $20/monthly software on the pile, because I just opted into Anthropic Pro to get access to their new “Opus” model in addition to some other goodies. This is some alpha-nerd stuff, so I’m going to my best to cut to the chase:
WHO: Anthropic is the brains behind Claude 3. They’ve been building AI that's not only smart but also ethically sound. So much so, it suddenly refused to role-play as characters; and once insulted my wife, calling her unethical for a dating show format she was using it for help with. Siiiick alignment.
WHAT: Anthropic has dropped three new AI models, each with their own unique skills. Sonnet is a blend of price and capabilities, while Haiku is designed like me: competent enough to be impressive, considering it is DIRT CHEAP.
But let's be real, Claude 3 Opus is the model everyone wants to swipe-right on…
BRING ON THE CLAUDE 3 OPUS: First off, Opus is a multi-talented powerhouse. It can handle complex tasks like analysis, coding, and creative writing.
Anthropic released a video demonstrating some Opus capabilities:
With this release, users can opt for the ideal combination of intelligence, speed, and cost to suit their use case.
Opus, our most intelligent model, achieves near-human comprehension capabilities. It can deftly handle open-ended prompts and tackle complex tasks.
— Anthropic (@AnthropicAI)
2:07 PM • Mar 4, 2024
And if you’re into benchmarks, here are some numbers Anthropic was throwing around, which compare Claude 3 models to OpenAIs GPT-4 and Google’s Gemini.
The Claude numbers are like, bigger. Which is better. ‘Cause, Texas. And ‘Murricah?
Other (early) benchmarks show Claude 3 Opus underperforming in things like this “emotional intelligence” test:
Source: gblazex on X
You know Gavin and I are going to dive deep into this and hopefully contextualize things for you on this week’s AI for Humans. But we’re also going to have Claude 3 Opus take center stage to DEBATE GPT-4 on a wide range of really stupid topics.
If you have questions or debate topics for Claude 3 (character TBA), feel free to fire ‘em our way!
And if you dig these daily drips, or there’s something you think we should feature/discss, let us know, and please consider sharing this with your friends and frienemies:
All aboard the Wednesday bus! See you lovelies then.
#KHBR
.k+g