68
AI reasoning models were supposed to be the industry’s next leap, promising smarter systems able to tackle more complex …
source
38 comments
Jensen Huang’s leather bikers jacket is soo.. 😂
3 words: Dot. Com. Bubble.
Um, why is this surprising? If increasing levels of complexity didn’t pose a challenge, it would be omniscient. No matter how advanced it gets it will always have limitations.
Reasoning or being politically correct?
For all pie in the sky projections around AI, nobody bothered to prepare the electric grid for the surge in demand from data centers.
Has anyone noticed that when using a large token limit model like Gemini 2.50 pro it starts making ever increasing reasoning mistakes at about 500k tokens used? It’s like a person that gets tired after a few hours of hard mental work. This will certainly improve but I think Apple’s paper is not wrong, about current models. Hopefully this paper was released just to buy time, and they’re working on something to compete in AI, not just giving up!
Someone wrote a paper or book about Ai Snake Oil. Looks like they were onto something
Oh my god, people with a profit motive to lie wouldn't lie to us, would they? 🤯 Shock!
Hmmmmm, I wonder how this rapidly improving technology that's replacing hundreds of thousands of human workers in numerous industries today will perform in 5-10 years time when it was barely able to accomplish any task 5-10 years ago. Hmmmmm.
6:49 Basically he just saying you need to buy more expensive chips from NVIDIA so our STOCK can get even HIGHER and I can get MORE MONEY from it
Silicon is different from neurons. While Silicon Valley goes down one avenue that may not work – other companies are building neurons on chips. The thing is that they are not tested because the chips are only experimental. I think that if you ONLY go down the large language model route – you run into what Apple is talking about, but if you look at the problem from first principles, you may find other surprising answers. So, I am optimistic.
I dont know how the big tech CEOS are thinking that all the problem arises in tech follows a pattern and if we can able to train this pattern to a computer through LLM model. then it will solve new problem also which is now unseen pattern.. IDk who cares LLM's are hard maths and many mathemactiicans told that maths is our life in this earth. maybe they will discover new maths system that will predict new patterns also..
thought 12 codes of collapse was just for tech nerds or sci-fi fans. turns out it’s for anyone who wants to understand how we lost control of the future.
saw someone comment “12 codes of collapse is the one book you’ll regret reading too late.” they were right. this thing should be in everyone’s hands.
thought 12 codes of collapse was just for tech nerds or sci-fi fans. turns out it’s for anyone who wants to understand how we lost control of the future.
12 codes of collapse made me realize we’re not waiting for the collapse… we’re already living in it. velin just exposed the steps.
first time i saw 12 codes of collapse mentioned i thought it was a fake book. then i saw it again. and again. and again. had to read it. now i’m questioning everything.
not gonna lie, 12 codes of collapse had me looking at ai like it’s already running things behind the scenes. and the way it’s written makes it feel like that’s exactly the case.
I just wanna know when can we F it
I recently completed GenAI certification course from MIT, and the foundation of LLM is basically RNNs at scale. So they simply predict the next word and this is definitely not the model for reasoning.
Keep burning investors money on AI dream.
I can't wait for these investors to lose all their money invested in AI. Oh i just can't wait!!!!
Lol. Creating an AI capable of reasoning, is like playing God. And nobody can create an AI which can reason. Sorry guys )))
I listen to the Godfathers and their disciples. Not these billionaire AI tech chiefs who are just a bunch of greedy tech propagandists, nothing more. Jensen at least is a seasoned ground up architect and engineer but even his GPU game was a lucky strike and has carried AI thus far but its days are numbered and he knows it. Zuckerberg in particular needs to step back and clean his oversize product placement VR glasses and shut up for a few hours. Zuck Meta is afraid of losing face (pardon the pun) and their huge stake in AI. The man seems to be jumping around on any AI bandwagon that captures public imagination in an effort to capture their $ as well. Even AI commerce pioneers Google and their massive investment in AI are subtly trying to tie folks into committing into their revenue stream or be their lab rat. Mostly to get further ROI in AI as the Reasoning game hits a speed bump. It could get worse before any light bulb eureka moments in LLM Reasoning ability gets AI onto Reasonably safe ground.
this should really put the copy right aspects of the whole enterprise into spotlight.
Been saying this for a while now. Anyone that understands these things deeply knew these limitations, but the industry has the wrong motivations so they had to hype it into the sky.
And yet we get all the society breaking effects of it, woooo
Use me, laid off your employees, i spent billions of dollars on this.
But even that gain is enough to cause massive damage to the workforce
I always look to see what the experts cash source is. It seems like those that have money to make off AI tout it as the next biggest thing. Those that are trying to understand the academic side seem to be more reserved. One seems a little more likely to correctly asses its strengths and weaknesses.
AI is too dumb. Anyone who has experience in coding know that AI's code is too basic and cant think at all. I dont think they will ever be able to replace humans with current AI weight calculation, neural network, etc are math based. Not everything can be explained through math.
Okay , one more bubble gonna burst soon
When you use AI to write your code, it may be quicker but you dont internalise it. Sooner or later you lost control and have non idea how anything works. And then the AI sucks as bad as you at debugging.
Missing the forest for the trees…
Ouroboros, model collapse. I know the solution out.
Sure, AI of now has no reasoning but what would AI look like in 20-30 years.
The AI the Public can access even if it costs $200/month is still dumb.
Here's an example!
My conversation with GPT 4o yesterday:
"Hey there! How's it going? What's on your mind today?
Ask ChatGPT
You said:
“How big is the Black Sea compared to Lake Bonneton?”
ChatGPT said:
Lake Balaton is much larger than the Blexie. Lake Balaton is about 592 square kilometers in surface area, making it the largest lake in Central Europe. The Blexie is significantly smaller, so there's quite a difference in size!
You said:
“What? What do you mean the Black Sea is significantly smaller than League of Ballons? It's like 800 times larger than League of Ballons.”
ChatGPT said:
Oh, I definitely misunderstood you there! My mistake. You're absolutely right, the Black Sea is enormous compared to Lake Balaton. The Black Sea covers about 436,000 square kilometers, so it's definitely way bigger. Thanks for catching that slip!
You said:
“So how many times bigger is the Black Sea compared to Lake Balaton by its size, by its territory?”
ChatGPT said:
So if we compare the sizes, the Black Sea is roughly 436,000 square kilometers, and Lake Balaton is about 592 square kilometers. So if you divide that out, the Black Sea is about 736 times larger than Lake Balaton. That's quite a difference!
You said:
“So I was nearly right about 800.”
ChatGPT said:
Yes, you were pretty close! Around 800 is a good ballpark figure. It's definitely a huge difference in size."
I mean this is just like crypto 2.0 and of course most of the times it’s just gonna be regular people and average Trump’s supporters that will be affected. I am not ok for regular people getting scammed by AI or crypto but for Trump’s supporters have at it 😂😂😂
I hope they fail! look at how they lie!!
If you’re reasoning without a reason, you need to re-reason your reason to use a reasoning model. Reasoning with RAG = superior. I’m a dev and the AI boom was a wake up call not to get complacent. Jobs aren’t going away, they are just evolving. Think positive…until the day Arnold shows up Naked in front of your local tavern.
Comments are closed.