Love it! Let’s go Daniel.! Yes it is a conspiracy. I had never heard of “Swagger Pop” before you pointed it out. My new favorite thing is to turn on closed captions for these ai launch videos. Very deep lyrics. “I made it. I made it hot..”
I thought for a sec you were going to do a parody of AI newsletters!
"Is using creative AI models trained on the work of human artists unethical? Does the answer change if we really want to see what a steampunk platypus looks like?
How will AI impact jobs, industries, and our entire way of life?
When is superintelligence coming, and how quickly will it kill us all?"
Excellent questions, person from 2022!
Also: o3 is rock solid. It still needs to improve, but it's kinda like a really smart high school kid now in a lot of ways.
Please, I'm a serious person. Why would I make fun of anything? Instead, I chose to focus on important matters like overanalyzing and mocking music tracks for AI launches.
And yes, o3 is pretty much my go-to ChatGPT model now for almost everything. It's helped me with lots of practical everyday stuff and is usually quite accurate for non-high-stakes decisions out of the box.
If you use the "Deep research" function in ChatGPT, I believe it's already always powered by o3 by default. But lately I prefer simply chatting to o3, because it returns answers faster than a "Deep research" report without a significant drop in quality of output.
I like that Deep Research will look at 150 different websites. I hate that the report is 40 pages long, but o3 can parse through and tell me the important stuff. It can’t just go out and do that on its own yet, at least not the way I’m using it. However, I understand (or I suspect, really) that o1 can do exactly this. I don’t wanna say I’m going to pull the trigger on that one, but I really might.
o3 should be a strictly better version of o1. It's a later model and it can use tools in addition to just reasoning. So if you're not quite getting what you want out of o3, it's not likely o1 will be better for your needs.
Here is where the dookie hits the fan as far as awful naming conventions that are hostile to the public. I am talking specifically about o1 pro mode, which you only get to use if you pay for the $200/month tier, and which you can only run five times per month (by contrast, o3 is limited to like 50-80 queries a week right now).
I can see why o3 has the lead in reasoning, but here's how ChatGPT puts it:
So is o1 Pro Mode “smarter”? Not across the board. But in its narrow domain — accuracy under pressure, long-chain reasoning, dense document parsing — it might still be the best model OpenAI’s ever released. Which is wild, because hardly anyone knows it exists.
I have a theory about this: it’s downstream of a trend started by Meta and Google. If you look back to 2022 and even late 2021, I suspect you’ll find ads like this for non-AI launches. It’s hard to make a splashy ad for software!
Thank you- great case for searching icon pop tracks from before and relaunching their creators for legend moods
"Hey, Google, find me a template AI launch soundtrack song, ASAP."
Love it! Let’s go Daniel.! Yes it is a conspiracy. I had never heard of “Swagger Pop” before you pointed it out. My new favorite thing is to turn on closed captions for these ai launch videos. Very deep lyrics. “I made it. I made it hot..”
To be fair, OpenAI o3 identified "Swagger Pop" as the genre. But hey, I'll take it!
And I think your "closed captions" idea might just set me off on yet another rabbit hole now!
The reality is a complete lack of creativity.
At every level.
Copycat videos for copycat products by copycat businesses.
That's a lot of cats.
Just wait until you see all the cat images these models can turn into cat videos, accompanied by a Suno song about cats!
I thought for a sec you were going to do a parody of AI newsletters!
"Is using creative AI models trained on the work of human artists unethical? Does the answer change if we really want to see what a steampunk platypus looks like?
How will AI impact jobs, industries, and our entire way of life?
When is superintelligence coming, and how quickly will it kill us all?"
Excellent questions, person from 2022!
Also: o3 is rock solid. It still needs to improve, but it's kinda like a really smart high school kid now in a lot of ways.
Please, I'm a serious person. Why would I make fun of anything? Instead, I chose to focus on important matters like overanalyzing and mocking music tracks for AI launches.
And yes, o3 is pretty much my go-to ChatGPT model now for almost everything. It's helped me with lots of practical everyday stuff and is usually quite accurate for non-high-stakes decisions out of the box.
Have you tried running a deep research report through o3, like analysis? Not a terrible combo.
If you use the "Deep research" function in ChatGPT, I believe it's already always powered by o3 by default. But lately I prefer simply chatting to o3, because it returns answers faster than a "Deep research" report without a significant drop in quality of output.
I like that Deep Research will look at 150 different websites. I hate that the report is 40 pages long, but o3 can parse through and tell me the important stuff. It can’t just go out and do that on its own yet, at least not the way I’m using it. However, I understand (or I suspect, really) that o1 can do exactly this. I don’t wanna say I’m going to pull the trigger on that one, but I really might.
o3 should be a strictly better version of o1. It's a later model and it can use tools in addition to just reasoning. So if you're not quite getting what you want out of o3, it's not likely o1 will be better for your needs.
EDIT: For reference, here's the LLM leaderboard: https://artificialanalysis.ai/leaderboards/models
Here is where the dookie hits the fan as far as awful naming conventions that are hostile to the public. I am talking specifically about o1 pro mode, which you only get to use if you pay for the $200/month tier, and which you can only run five times per month (by contrast, o3 is limited to like 50-80 queries a week right now).
I can see why o3 has the lead in reasoning, but here's how ChatGPT puts it:
So is o1 Pro Mode “smarter”? Not across the board. But in its narrow domain — accuracy under pressure, long-chain reasoning, dense document parsing — it might still be the best model OpenAI’s ever released. Which is wild, because hardly anyone knows it exists.
^ that's the model I am eager to try out
It's funny when you realize it and you see how these patterns manifest and then vanish.
Yet another proof that we live in the Matrix!
I have a theory about this: it’s downstream of a trend started by Meta and Google. If you look back to 2022 and even late 2021, I suspect you’ll find ads like this for non-AI launches. It’s hard to make a splashy ad for software!
Oh, so the rabbit hole goes deeper than AI. Is there truly no end to this madness?!
Haha. Brilliant Daniel! I can’t stop laughing 😂 I never noticed this as I mostly watch with the sound off.
Happy to introduce you to the mad world of AI launch tracks!