o3 doesn’t seem that groundbreaking to me. It seems iterative. 

We’ve known that computers are “smarter” than us. Since the 80s. That’s why we use them. 

Let me know when these models show *intention* and *creativity*. 

That’s what makes us different. 

Until then they remain really, really, really good word predictors.

leader • reader • taco eater • fly eagles fly • @bunches • /whiskey • /nativefun • previously addepar, linkedin • https://nf.td/derek

Disclaimer: I’m willing to eat these words.

Seems to me we keep moving the goal post. 

If we were told in 90's an algorithm read a bunch of books & crunched down scraped data, learning itself how to code and (poorly) do arithmetic people would be loosing our minds.

Agreed they’d lose their minds.  

Two things can be true: be amazed and call it iterative.

I’m extraordinarily bullish AI, but intention and creativity have always been my personal goal post for calling something AGI or sentient.

What we have today is an amazing and groundbreaking 10000x iteration on the interface and capabilities of the original Eliza, Encarta, Quickbasic, and the calculator. 

No sarcasm there. It is amazing. But it’s not AGI. 

That said, it’s completely possible we do have models that show independent intentionality but companies like Anthropic and OpenAI don’t release them for safety’s sake. 

Which I’d completely understand (and be grateful for) as well.

They definitely “feel” the same. I do think they have some new use cases because of the speed and increased complexity. 

I use them daily and they do increase my productivity. 

But the hallucinations and wrong information are great reminders of their shortcomings. 

Until I ask it “where do you want this conversation to go?” And it responds with something other than a version of “I’m just here to be helpful”, I’m not wowed.

well said and agree. they all “feel” the same to me after a bit of use. hollow repackaging of human intellectual goods, averaged, smoothed, and banal. Useful for trite timewasters, otherwise *just a computer.*

Search engines were better encyclopedias.

LLMs are better search engines.

That’s not a bad thing, it’s just not what a lot people think it is (AGI), and it can’t do the things they expect it to do.

Also we know o1 is best at deceiving the user with wrong information. O3 will probably excel at this

Well i have a fren that loves to play flamenco and it’s a bot. A lovely fella