visarga

26 ms·

631.
by visarga 1y ago
> their dream of the perfect slave machine I don't get if AI is supposed to be a slave or a machine. Is it sentient or a toaster?
632.
by visarga 1y ago
Essentially, instead of modifying the prompt itself, the system intelligently directs the prompt to the LLM that is best suited to handle it based on its learned performance and efficiency characteristics for similar types of queries. It&#x
633.
by visarga 1y ago
It's ok if you can update the router over time, the more data you have the better.
634.
by visarga 1y ago
They are a respectable institution, they only lie about WMDs in order to justify war and death. That is why we need to protect their IP, to preserve their activities in the future.
635.
by visarga 1y ago
> I have struggled to get any of the RAG approaches to handle this effectively. You need to annotate your text chunks. For example you can use a LLM to look over the chunks and their dates and generate metadata like summary or entities.
636.
by visarga 1y ago
I would rather only see websites without ads, or that explicitly allow ad blocking. Websites with ads are shit anyway, clickbait and slop all the way down. The whole internet has been enshittified by the battles for user attention. Nothing
637.
by visarga 1y ago
This law is based on the idea that a restaurant patron should be forced to eat everything on their plate, or else it is infringing on the IP of the cook. "Sir, we made the salad with onions and you skipped them, we need to forcefully s
638.
by visarga 1y ago
If you use Claude to generate apps in the artifact they are very similar - self contained and easy to create and share.
639.
by visarga 1y ago
The crucial information is missing - accuracy comparison with other OCR providers. From my experience LLM based OCR might misread the layout and hallucinate values, it is very subtle but sometimes critically wrong. Classical OCR has more pr
640.
by visarga 1y ago
> Things that were previously overwhelming are now almost trivial. Not if you have to stake your career on them. AI takes no liability, it is all yours
641.
by visarga 1y ago
> Pretty much, make a separate LLM request to check the query against the existing context to see if more info is needed, and a second check after generation to see if output text exceeded it's knowledge base. They are unreliable at
642.
by visarga 1y ago
Verification is the bottleneck, not ideation. LLMs can generate anything on tap, but solving any non-trivial problem requires iteration between thinking, doing and observing outcomes. The real world is too complex to be simulated by AI or h
643.
by visarga 1y ago
> I think the most sensible answer would be something like UBI. Having had the experience of living under communist regime prior to 1989 I have zero trust in the state providing support, while I am totally dependent and have no recourse.
644.
by visarga 1y ago
> displace humans ... AI can displace human work but not human accountability. It has no skin and faces no consequences. > can be trained to the new job opportunities more easily ... Are we talking about AI that always needs trainers
645.
by visarga 1y ago
Output text is word salad every few words apart. You can't scale n-gram counting enough to make it work.
646.
by visarga 1y ago
goats have too many parameters, they are like GPT-4
647.
by visarga 1y ago
> I'm starting to think this is a deeper problem with LLMs that will be hard to solve with stylistic changes. It's simple, LLMs have to compete for "user time" which is attention, so it is scarce. Whatever gets them m
648.
by visarga 1y ago
> algorithms just take other people's work, mix it so that any individual training input is unrecognizable and resell it back to them LLMs are huge and need special hardware to run. Cloud providers underprice even local hosting. Man
649.
by visarga 1y ago
> In human learning we do this process by generating expectations ahead of time and registering surprise or doubt when those expectations are not met. Why did you give up on this idea. Use it - we can get closer to truth in time, it take
650.
by visarga 1y ago
> They don't really know if they're right. Neither do humans who have no access to validate what they are saying. Validation doesn't come from the brain, maybe except in math. That is why we have ideate-validate as the cor
651.
by visarga 1y ago
There is no Truth, only ideas that stood the test of time. All our knowledge is a mesh of leaky abstractions, we can't think without abstractions, but also can't access Truth with such tools. How would Truth be expressed in such a
652.
by visarga 1y ago
And loads of money to pay for tokens, because every month I am out of tokens after a week or two.
653.
by visarga 1y ago
It sees tokens not letters like us. And has to recite tokens in reverse order, and their letters in reverse order, over a set of 200K tokens. Token codes are arbitrary numbers associated with word fragments, they convey no letters.
654.
by visarga 1y ago
Yes, for information and reference we already had Wikipedia and billions of web pages indexed in Google, searchable by keyword. For questions we had reddit, StackOverflow and forums. For chatting we had social networks, chatting with real h
655.
by visarga 1y ago
Let's change this game a bit. Spell "understanding" in your head in reverse order without spending twice more time than forward mode. Can you? I can't. Does that mean we don't really understand even simple spelling?
656.
by visarga 1y ago
Understanding is always functional, we don't study medicine before going to the doctor, we trust the expert. Like that we do with almost every topic or system. How do you "understand" a company or a complex technological or b
657.
by visarga 1y ago
I tried GPT-5 in Cursor today and got the same thing - it started reading and thinking and thinking, and it was quite repetitive. It was unexpected. I wrote it off as "maybe the Cursor GPT-5 prompts are not refined yet".
658.
by visarga 1y ago
> This is very dumb. Of course you can. When it's your problem being delegated, you can't delegate consequences away. I can eat for you, but you won't get satiated this way.
659.
by visarga 1y ago
Which means real experience still takes years. But you need to consider the speed at which coding agents improve. Maybe next year they will be more reliable to use without domain experience. Today? You can get a small app or POC without kno
660.
by visarga 1y ago
> If that’s not vibe-coding, then what is it? Blind-coding.

More ›