visarga

42 ms·

721.
by visarga 1y ago
> My tolerance for AI style "knowledge" is lower and lower every day. We like to think humans possess genuine knowledge while AI only learns patterns. But in reality do we learn medicine before going to the doctor? or do we eng
722.
by visarga 1y ago
To insist it's "just code and math" is to fundamentally mistake the rules for the game. Knowing the rules of Go without playing grants no insight into strategy, just as showing someone the code for Conway's Game of Life
723.
by visarga 1y ago
"The Illusion of Thinking" is bad, but so is the opposite "it's just math and code". They might not be reasoning like humans, but they are not reducible to just code and math either. They do something new, something
724.
by visarga 1y ago
Why link to this copyright maximalist site with paywall? They argued against copyright in 2001 (NYT vs Tasini) when it was in their favor, and recently pro copyright (NYT vs OpenAI) because it was in their ..favor. They want to put restrict
725.
by visarga 1y ago
Maybe Apple thought they need to make a desperate move to deflect from their lack of AI?
726.
by visarga 1y ago
If they really want to improve MacOS they should add synonym based search in Settings. I should not be forced to know by heart the arcane speech their marketing department chose to replace consecrated terms. Just today it took me 5 minutes
727.
by visarga 1y ago
I guess LLM and AI are forbidden words in Apple language. They do their utmost to avoid these words.
728.
by visarga 1y ago
> prompt engineers point at them as evidence of the fast-approaching software engineer obsolescence Maybe journalists and bloggers angling for attention do it, prompt engineers are too aware of the limitations of prompting to do that.
729.
by visarga 1y ago
The idea is good, but we should commit both documentation and tests. They allow regenerating the code at will.
730.
by visarga 1y ago
> when you drill down to first principles It is a trap to consider that first principles perspective is sufficient. Like, if I tell you before 1980: "iterate over f(z) = z^2 + c" there is no way you are going to guess fractals
731.
by visarga 1y ago
I am interested in TTS for reading web pages and LLM responses but it's too expensive. At this price point I can't look at it. I will continue using local TTS, not as great but instant, allows tracking text as it read it and works
732.
by visarga 1y ago
Copyright is not going well. The rights of millions of people are trampled by companies, both the content we post on social networks and our private AI chats. Our voice doesn't matter. Copyright was supposed to protect expression and k
733.
by visarga 1y ago
NYT wants it both ways. When they were the ones putting freelancer articles into a database to rent, they argued against enforcing copyright and for supporting the new industry, and that it was too hard to revert their original assumptions.
734.
by visarga 1y ago
And they are doing this over literally "old news". Expired for years, of no value.
735.
by visarga 1y ago
> If there is no intention behind the words why would it be interesting to talk to them in the first place. But of course there is a story behind them.
736.
by visarga 1y ago
I recognize some GPTisms in this article. "Holistic Thinking" and "Ethical Considerations" being things humans are necessary for... it likes to say that a lot. I don't agree much, it seems superficial LLM limitation
737.
by visarga 1y ago
All right, that does it! I've had enough internet for today.
738.
by visarga 1y ago
Brains don't have innate grammar more than languages are selected to fit baby brains. Chomsky got it backwards, languages co-evolved with human brains to fit our capacities and needs. If a language is not useful or can't be learne
739.
by visarga 1y ago
A billion people use LLMs which have been trained to refuse unethical demands. LLMs show higher than human average ethical thinking. So it's not like a lawnmower by any stretch.
740.
by visarga 1y ago
> The LLM market is a market of hype It's certainly not magic or super human, but it's not nothing either. Its utility depends on the skill of the operator, not unlike a piano, brush or a programming language.
741.
by visarga 1y ago
> What jobs will AI create that AI cannot itself do? AI lack skin-in-the-game, they cannot bear responsibility for outcomes. They also don't experience desires or needs, so they depend on humans for that too. To make an AI useful yo
742.
by visarga 1y ago
> Because an LLM is an unthinking machine. Unthinking can be pretty powerful these days.
743.
by visarga 1y ago
Humans aren't good at validation either. We need tools, experiments, labs. Unproven ideas are a dime a dozen. Remember the hoopla about room temperature superconductivity? The real source of validation is external consequences.
744.
by visarga 1y ago
I too have sat at pianos and banged the keys but no valuable music came out. It must be because the piano is a bad instrument. /s
745.
by visarga 1y ago
> An 8GB Macbook could use Gemma 3 4B at 2.5GB+1GB, but this is probably not worth doing. I am currently using this model on a Macbook with 16GB ram, it is hooked up with a chrome extension that extracts text from webpages and logs to a
746.
by visarga 1y ago
Spoken like someone who has experience with unstructured information extraction, for sure.
747.
by visarga 1y ago
Maybe we just need models that can reference spans by start:end range. Then they can pass arguments by reference instead of explicit quotation. We can use these spans as answers in extractive QA tasks, or as arguments for a code block, or t
748.
by visarga 1y ago
> Regardless of the mechanism, the foundational 'conceit' of LLMs is that by dumping enough syntax (and only syntax) into a sufficiently complex system, the semantics can be induced to emerge. Syntax has dual aspect. It is both
749.
by visarga 1y ago
My favorite argument against SP is zero shot translation. The model learns Japanese-English and Swahili-English and then can translate Japanese-Swahili directly. That shows something more than simple pattern matching happens inside. Besides
750.
by visarga 1y ago
It would be a mistake to take the banality of current LLM outputs and extrapolate that into the future. Of course they are going to get better. But that is not the point - it is that in the chat room the human and LLM spark ideas off each o

More ›