lukev

23 ms·

421.
by lukev 2y ago
This is a pretty big technical achievement, and I am excited to see this type of advancement in the field. However, I am very worried about the utility of this tool given that it (like all LLMs) is still prone to hallucination. Exactly who
422.
Ask HN: How do you organize work as a solo developer?
61 points by lukev 2y ago|42 comments
423.
by lukev 2y ago
Unless it’s bloat in the military or corporate welfare , or big-brothering what people do with their bodies.
424.
by lukev 2y ago
Hence the other aspects of antitrust law. Albeit not enforced well of late.
425.
by lukev 2y ago
Or one of them could lower to 4.35 and take all the business. Which is normally what would happen without collusion.
426.
by lukev 2y ago
Well, this isn't the "End of Truth" or anything like that. But it's certainly going to change how people relate to photos. Yes, you could always photoshop something but that takes both knowledge and effort. This takes ne
427.
by lukev 2y ago
What are you talking about? Efforts to make it more difficult to vote are almost exclusively right wing. Just look at the list of cases tracked by Democracy Docket: https://www.democracydocket.com/cases/
428.
by lukev 2y ago
Yes. Price fixing can happen in any market where participants on one side of the market collude to set prices. It is mostly illegal in the US: https://en.wikipedia.org/wiki/Price_fixing#United_States
429.
by lukev 2y ago
Some quotes from the filing: > Discussing a different RealPage product, another landlord said: “I always liked this product because your algorithm uses proprietary data from other subscribers to suggest rents and term. That’s classic pri
430.
by lukev 2y ago
They are _used_ very extensively, but not in web pages. Many industry domains rely heavily on linked data and ontologies. It's true that they are less commonly embedded as semantic data in web pages. There's a real bootstrapping p
431.
by lukev 2y ago
That's only schema.org! Linked data is so much bigger than that. Many ontologies have a "poem" type (for example dbpedia ( https://dbpedia.org/ontology/Poem ) has one), as well as other publishing or book-
432.
by lukev 2y ago
Really? Trump has always been much weaker in diplomacy with nations that are historically antagonistic to the US. He'll have a meeting with their leader, let them flatter him, and come away convinced that they're "very nice&q
433.
by lukev 2y ago
Yes, I'm quite sure the CTO of a leading observability platform is simply confused about terminology.
434.
by lukev 2y ago
The MBA-brained focus on short term profits is a relatively recent development and was not nearly as prevalent back then. Safety engineering is about as long-term a value as you can imagine, so it makes sense that it's one of the thing
435.
by lukev 2y ago
This is also pretty easy with nothing but SSH, setting up a tunnel through any server to which you have SSH access (including extremely cheap/free ones.)
436.
by lukev 2y ago
> LLMs are discriminating actual truth No. They are functions of their training data. There is absolutely no part of a LLM that functions as a truth oracle. If training data contains multiple conflicting perspectives on a topic, the LLM
437.
by lukev 2y ago
It's rapidly getting to the point where the cure is worse than the disease, when it comes to this kind of product.
438.
by lukev 2y ago
Nuance: they aren't sensitive to "truth", they are sensitive to recurring or consistent input in the training data.
439.
by lukev 2y ago
You could argue it's 1d, actually, since sequence is fundamental, not positioning on the x axis. At any rate it's (mostly+) categorically different from what visual programming attempts. Code must be read, comprehended, and a ment
440.
by lukev 2y ago
I got the Apple Vision Pro with the hope to tinker with such things. Is one more dimension enough to "unlock" visual programming? I don't know, and unfortunately not many seem interested in exploring it.
441.
by lukev 2y ago
> just as easy Citation needed. The social context matters. Alcohol and caffeine are not a wise combination. Socially, consuming both in parallel isn't common (Red Bull excepted, I suppose, but that was exactly the target market and
442.
by lukev 2y ago
Counterpoint: if I'm already sitting at my computer, I will 100% use my computer to write texts and make calls. I agree with the gp. The only benefit of a phone is portability and ease of access.
443.
by lukev 2y ago
I think the average consumer does notice the amount of crap that pops up when they boot their brand new computer, and gets frustrated when they can't remove it easily.
444.
by lukev 2y ago
I am by no means a vegetarian, and I enjoy a good steak or burger very much. I hope the possibility of accessing them never goes away. But it seems pretty obvious that we need to shift our consumption habits towards more sustainable foods.
445.
by lukev 2y ago
This is one of the key tenets of a universal basic income program: it's universal, without means testing. Every citizen gets it, no questions asked. This drastically reduces both the cost of administering the program and improves the e
446.
by lukev 2y ago
Yes, I think everyone understands that leaks happen. The problem being pointed out here is that you can't leak what you don't collect: this is a lot more data, and a lot more sensitive than what MS has ever collected before.
447.
by lukev 2y ago
Well, that's the great question. If there's enough of the "essence of reason" represented in human text, then it's possible in theory some sufficiently large LLM could eventually could grok it and fully generalize o
448.
by lukev 2y ago
One simple answer that I don't see people give often to this question is: yes, but only the extent which reasoning is reflected in language. This explanation pretty much tracks exactly with model performance on reasoning tasks. Think a
449.
by lukev 2y ago
This is a good post, but it's somewhat myopically focused on typical "business" data. The most interesting applications for "big data" are all (IMO) in the scientific computing space. Yeah, your e-commerce business
450.
by lukev 2y ago
A relevant point to this is the notion of "System-1" vs "System-2" thinking. Somewhat dubious when applied to actual human psychology but I think a valid metaphor for how LLMs work: they are only capable of System-1 thin

More ›