visarga

17 ms·

421.
by visarga 10mo ago
The stochastic parrot framing makes some assumptions, one of them being that LLMs generate from minimal input prompts, like "tell me about Transformers" or "draw a cute dog". But when input provides substantial entropy o
422.
by visarga 10mo ago
I don't think it will become well rounded because that is not cost sensitive. Intelligence is sensitive to cost, it is the core constraint shaping it. Any action has a cost - energy, materials, time, opportunity or social. Intelligence
423.
by visarga 10mo ago
Coding agents are going to become better and used everywhere, why train for the artisanal coding style of 2010 when you are closer to 2030? What you need to know is how to break complex projects in small parts, improve testing, organize wor
424.
by visarga 10mo ago
AI helps ship faster but we need to code 1.7x more tests (with AI) and it's allright
425.
by visarga 10mo ago
It will come around, we have rudimentary computer use agents and ability to record UIs for LLM agents. They will me refined and the agent can test UIs as well. For UIs I do a different trick - live diagnostic tests - I ask the agent to writ
426.
by visarga 10mo ago
I agree with the author overall. Manual testing is what I call "vibe testing" and I think by itself is insufficient, no matter if you or the agent wrote the code. If you build your tests well, using the coding agent becomes smooth
427.
by visarga 10mo ago
I do. Gave up using Gemini directly.
428.
by visarga 10mo ago
Having a comprehensive spec and test suite is an absolute requirement, without it all you got is vibe-testing, LGTM feels. As shown by the OP, you can throw away the code and regenerate it back from tests and specs. Our old manual code is n
429.
by visarga 10mo ago
I think the only non-slop parts of the web are: open source, wikipedia, arXiv, some game worlds and social network comments in well behaved/moderated communities. What do they share in common? They all allow building on top, they are s
430.
by visarga 10mo ago
I think specs + tests are the new source of truth, code is disposable and rebuildable. A well tested project is reliable both for humans and AI, a badly tested one is bad for both. When we don't test well I call it "vibe testing,
431.
by visarga 10mo ago
If I were a creative I would avoid seeing any work I am not legally allowed to get inspired by, why install furniture into my brain I can't sit on? I see this kind of IP protection as poisoned grounds, can't do anything on top of
432.
by visarga 10mo ago
>Red flag. In other words you don’t understand the implementation well enough to know if the AI has done a good job. Red flag again! If your protection is to "understand the implementation" it means buggy code. What makes a cod
433.
by visarga 10mo ago
After 4 hours of vibe coding I feel as tired as a full day of manual coding. The speed can be too much. If I only use it for a few minutes or an hour, it feels energising.
434.
by visarga 10mo ago
Shorts are everywhere because it is the most addictive form of media, easy to consume, no effort required to follow through. More generally I think the problems we got into were inevitable. They are the result of platforms optimizing for th
435.
by visarga 10mo ago
> The issue is whether they're paying us for taking the human content of the last two thousand years and baking it into their generators. Payment for content access is a sure way to limit progress and freedom. Should I pay you based
436.
by visarga 10mo ago
May I point out that we don't know in detail how most code runs? Not talking about assembly, I am talking about edge cases, instabilities, etc. We know the happy path and a bit around it. All complex systems based on code are unpredict
437.
by visarga 10mo ago
I think it is AI worded, but ideas come from a real human.
438.
by visarga 10mo ago
Just write 3x more tests and let the agent go. It's simple.
439.
by visarga 10mo ago
That is a good approach, bottom up, manage complexity. But the general picture is - you set the direction and hold the model responsible, it does the actual work. Think of it as your work is the negative of the AI work, it writes the code,
440.
by visarga 10mo ago
> It's not what can humans provide for each-other, it's what can humans provide for a handful of ultra-wealthy oligarchs. You can definitely use AI and automation to help yourself and your family/community rather than the
441.
by visarga 10mo ago
> What happens when there are no more hosts to donate more training-blood? LLMs have over 1B users and exchange over 1T tokens with us per day. We put them through all conceivable tasks and provide support for completing those tasks, an
442.
by visarga 10mo ago
An ATM is a reliable machine with a bounded risk - the money inside - while an AI agent could steer your company into bankruptcy and have no liability for it. AI has no skin and depending on application, much higher upper bound for damage.
443.
by visarga 11mo ago
Not pictures with this composition, same jacket, etc - yes, there are images but they are different, this fits like a key in the lock to the original
444.
by visarga 11mo ago
> the interpretability research from Anthropic [1] suggests that structures corresponding to meaning do exist inside those bundles of numbers and that there are signs of activity within those bundles of numbers that seem analogous to tho
445.
by visarga 11mo ago
The loop, or more precisely the "search" does the novel part in thinking, the brain is just optimizing this process. Evolution could manage with the simplest model - copying with occasional errors, and in one run it made everyone
446.
by visarga 11mo ago
Sentence completion does not give it justice, when I can ask a LLM to refactor my repo and come back half an hour later to see the deed done.
447.
by visarga 11mo ago
You can make a separate model for the task, which is based on well chosen features and calibrated from actual data. Then the LLM only needs to generate the arguments to this model (extract those features from messages) and call it like a MC
448.
by visarga 11mo ago
Well, yes because thinking soon requires interacting, not just ideating. It's in the dialogue between ideation and interaction that we make our discoveries.
449.
by visarga 11mo ago
Not always, you can improve the loop by putting something real inside, like, a code execution tool, a search engine, a human, other AIs or an API. As long as the model can make use of that external environment its data can improve. By the s
450.
by visarga 11mo ago
Yesterday I stumbled onto a well written comment on reddit, it was a bit contrarian, but good. Then I was curious and looked at their comment history and found it was a one month old account with many comments of similar length and structur

More ›