Not arguing against 'great', but cost efficiency is questionable. for 10% you can get two used 3090. The good thing about LLMs is they are sequential and should be easily parallelized. Model can be split in several sub-models, by the number of GPUs. Then 2,3,4.. GPUs should improve performance proportionally on big batches, and make it possible to run bigger model on low end hardware.
it may be even cheaper to use big models through API. Claude, GPT, whatever. Rental is efficient only for big butches, while API is priced per call/size and is cheap for small models.
It's a trick. If you donate something to charity you can write it off the taxes. After IPOs most new billionaires did it, including Zuk. It's enough to just promise it publicly. There is no time limit, it can take years to 'donate', but writing off is immediate. To optimize it even further they create their own dummy charities full of friends and relatives.
Simple tax optimization. Like new billionaires promising to significant donations. a) they don't have to donate. b) they can immediately slash those promised donations from taxes.
I'm not an expert, but I'll do it for learning. Then open source if it works. As far as I understand this approach requires a vector database and LLM which doesn't have to be big. Technically it can be implemented as local web server. Should be easy to use, just type and get a sorted by relevance list.
This is cool, but how about local semantic search through tens of thousands articles and books. Sure I'm not the first, there should be some tools already.
> it will just have much better precision than us.
and much faster with the right hardware. And that's enough if AI can do in seconds what humans takes years. With o3 the price is only the limit, looks like.
After collision drone should shut itself down and drop like a rock. Looks like it wasn't the case here. And there should be nobody on the ground below, of course. Flying over that lake would be safe
With this a lot of damage can be done even without drone. As for weaponized it's not a future, it's a reality in Ukraine for years now. Defense against them is difficult to impossible. A bodyguard who can sacrifice himself may sometimes work.
They are not mutually exclusive. Likely we'll get more clear separation of architecture and underlying technology. In this case agents (i.e. architecture) can use different technologies or mix of them. Including 'AI' and algorithms. The trick is to make them work together.
> AGI is special. Because one day AI can start improving itself autonomously
AGI can be sub-human, right? That's probably how it will start. The question will be is it already AGI or not yet, i.e. where to set the boundary. So, at first that will be humans improving AGI, but then... I'm afraid it can get so much better that humans will be literally like macaques in comparison.
> if the underlying LLMs dont get any better, there is no reason to expect the system built out of them to get any better.
Actually o1, o3 are doing exactly this, and very well. I.e. explicitly: by proper orchestration the same LLM can do much better job. There is a price, but...
> you would expect to be able to build agentic systems out of much smaller LLMs
Good point, it should be possible to do it on a high-end pc or even embedded.
That's a great speed. I'll never reach because I have to read some paragraphs several times before I understand the article. Some times using pen a paper...
Anyway, you can still improve you speed with the same level of understanding by asking chatgpt for summary.
> multiple years in with much investment and little return
Hmm.. where do you live? It's quite transforming already in many areas and not going to stop any time soon. I'll call it 'The Last Explosion', as opposite to 'AI Winter'. By that I mean this explosion with result in AGI. Likely sub-human first, then super-human.
from https://www.asacomputers.com/nvidia-l40s-48gb-graphics-card....
nvidia l40s 48gb graphics card Our price: $7,569.10*
Not arguing against 'great', but cost efficiency is questionable. for 10% you can get two used 3090. The good thing about LLMs is they are sequential and should be easily parallelized. Model can be split in several sub-models, by the number of GPUs. Then 2,3,4.. GPUs should improve performance proportionally on big batches, and make it possible to run bigger model on low end hardware.