my latest on importance of AI and the stupidity of countries that line up behind America in this AI battle without any effort to help their own countries.
Artificial Intelligence thread
- Thread starter 9dashline
- Start date
how are you still posting links here without converting them into twitter.com?
One more warning and you take a 1 month vacation from this thread
AI is certainly going to be very useful. But wasting billions running GPUS to train glorified Chatbots is not going to be the next big thing. People thought completely driverless cars will be mainstream 5 years ago. But it turned out, training neural networks to think and take into account every kind of scenario is not so easy. Moreover, people themselves will never trust full self-driving anytime soon. One accident and the trust is all gone.
Glorified Chatbots pretty much the same thing. Yes, they are making people's lives easier by making it easy to search and generate text and code. but I doubt they can replace one actual human anytime soon. There is just too much caveat and nuance in the real world that needs actual human to actually do things. Chatbots can make things easy, but they can't replace actual decision makers and doers.
Moreover, even this making life easier part is not monetizable due to advancements in AI efficiency, free AI models will always be as good as the paid ones, so ultimately no one will make any money out of the chatbots.
Its actually good that China is being throttled by sanctions not to sink billions into training chat bots. I expect we will see an AI winter again very soon and subsequent collapse in the AI bubble in the US. China on the other hand is focusing more on practical AI and that is here to stay.
As usual, OpenAI is paying people to solve complex math problems and then training against those solutions so that it can score better on its reasoning model o3 and claim some super performance.
The company is run by an absolute scam artist in Sam Altman who has a super talent in just getting people to overpay whatever he has to sell.
So, be suspicious of whatever OpenAI claims in the future. It has no integrity at all.
Deepseek r1 weights released...
This is like dropping an atomic warhead on OpenAI
Mother of God this one is stronk... beats o1 medium....
Last edited:
And o1 is their best model. I recently used o3 for staged reasoning, and the outputs are arguably worse.
only 1.52kb wow. Arduino Nano is the new H100.
I knew the XHS translations were almost too good. Had to be LLM.
AI winter (or at least semi-winter) already happened for most startups a couple years ago when people realised how overhyped current AI technologies are. There is still very good research happening and the area is still an investment hotbed but the days of being able to put 'AI' in your pitch deck and getting millions are over.
a distill of R1, in every size is here...
Sam Altman on suicide watch
Attachments
-
Screenshot_2025-01-20-07-57-24-25_cbf47468f7ecfbd8ebcc46bf9cc626da.jpg