Ukv

63 ms·

781.
by Ukv 3y ago
Also the wrong way to implement an expiry data, since it'd work fine again when the day goes below 21 or month below 11, even if the year is 2021 or greater - which seems to be what happened if they only noticed it in November 2022 rat
782.
by Ukv 3y ago
> They can believe whatever they want, that doesn't mean I have to share their beliefs. I don't claim you have to, and noted that whether you're willing to make such an inference depends on the views you subscribe to. >
783.
by Ukv 3y ago
> And that has nothing to do with "what philosophy of mind you subscribe to", There are definitely those that would reason "Using the tools and expertise available to us, this behaves indistinguishably from a human - there
784.
by Ukv 3y ago
> It doesn't test "Can a machine think" so much as it tests "Can a human be tricked". But importantly, Turing didn't just mean it'd be tested on ability to make small talk. He chose the format because h
785.
by Ukv 3y ago
True that A* would later go on to be used in computer game bots (as LLMs may eventually be) where "an AI" doesn't necessarily correspond to the field of AI, but A* was still within what was understood to be the field of AI at
786.
by Ukv 3y ago
> it's Gary Marcus. He's an academic with a lot of prestige to lose if the LLM approach is actually good/useful/insightful I wouldn't bet on loss of perceived expert on the other side status or prestige no matter
787.
by Ukv 3y ago
When it was published, A* was considered within the field of AI. The paper[0]'s authors were from Stanford's AI group. The "AI effect"[1] describes how the field tends to get left with only the cutting-edge research. [0]
788.
by Ukv 3y ago
> "May have died" > Do they have any actual evidence to this? The closest thing I saw in the article is [...] > Do they have any evidence of that many deaths? All I am seeing is "estimates". The article is about
789.
by Ukv 3y ago
Do any github.io pages load for you? Might be that they got reports of a github.io page being used to link to malware, and decided to block the entire domain.
790.
by Ukv 3y ago
Definitely short, but I feel it is still ~80% marketing. You could cut out: > As you saw at Microsoft Ignite this week, we’re continuing to rapidly innovate for this era of AI, with over 100 announcements across the full tech stack from
791.
by Ukv 3y ago
> given that there is a story on the front page of HN about once a week about why open-source maintainers should be paid and how the current model isn't sustainable Opinion articles on HN notwithstanding, do you not think that free
792.
by Ukv 3y ago
> I don't see copyright or a record label as artificial scarcity It's allocating resources towards preventing distribution. Maybe you'd agree that it's "direct" artificial scarcity, but believe without that
793.
by Ukv 3y ago
nurple said: "their creation is held behind legal walls in order to create, by definition, an artifical scarcity from which it derives its monetary value" In my example "their creation" is a textbook, and "legal wal
794.
by Ukv 3y ago
> It's literally limited by the hours in the day If a textbook isn't available because it's still being written by the author, then true, that's the real scarcity of the author's time. But if a textbook is finish
795.
by Ukv 3y ago
Could work for spear-phishing, or impersonating a widely-known trusted figure. I can't really see it working for cold-calls that pretend to be someone the victim knows (like the terrifying ransom calls), since the operations work at a
796.
by Ukv 3y ago
> one of the factors affecting whether the act of copying is fair use, according to Congress, is “the effect of the use upon the potential market for or value of the copyrighted work”. Today’s generative AI models can clearly be used to
797.
by Ukv 3y ago
> There are currently no versions of reverse image search existing today that are as good as what Google was offering several years ago. Which have you tried? Finding unwatermarked versions of stock image photos is a pretty low bar that
798.
by Ukv 3y ago
For reverse image search, Google was massively behind competitors (especially Yandex) for a long time - though the gap has narrowed with Google Lens. For regular search results, I often find myself frustrated that Google seems to prefers ad
799.
by Ukv 3y ago
To my understanding this is a small low-energy chip designed for applications like basic object detection in a doorbell camera, or speech recognition for a home assistant. I think it's a bit misleading to say "faster [...] than Nv
800.
by Ukv 3y ago
For integer and floating-point operations, but not for fixed-point operations which are neither.
801.
by Ukv 3y ago
> One requires reporting of any model trained using more than 10^26 integer or floating point operations total On an unrelated note, OpenAI announces GPT5 will be trained with fixed-point arithmetic.
802.
by Ukv 3y ago
> If I take a copyrighted song and remix it do I owe the artist a royalty? Yes. I don't think anyone's saying that the output of the model should be subject to more lenient copyright standards than human creations. If you&#x27
803.
by Ukv 3y ago
> in a way that is completely dependent upon their own prior work Ultimately information has to come from somewhere. If something has no information about what a "car" is, it cannot paint a car more successfully than a random g
804.
by Ukv 3y ago
Making "style" into private property subject to infringement seems like the far more slippery slope to me. I think existing copyright laws are more than sufficient, for cases where there are actually generations with substantial s
805.
by Ukv 3y ago
You likely benefit from machine learning applications constantly without realizing. The spam filters for your email, the scanning for defects of the products you use and the rails they were delivered on, when you enter a search query or tra
806.
by Ukv 3y ago
> but we don't need to store the output to recreate it; we need only store the formula or code. This doesn't exempt it from information theory. If a 3GiB model was trained on 5 billion images, it fundamentally cannot have more
807.
by Ukv 3y ago
True that ML is good for lossy compression, but it's not "produce a copy of any image with less than a byte average of information" good, like is claimed in the Joseph Saveri Law Firm lawsuit against Stable Diffusion: > Be
808.
by Ukv 3y ago
To my understanding, trademark law should already "protect their identity" by preventing use of their logo, and copyright law would already prohibit an "intentional and exact clone". A patent differs in that it counts as
809.
by Ukv 3y ago
It can be how the law works; like in Field v. Google Inc. for hosting a cached copy of a site: "Google reasonably interpreted absence of meta-tags as permission to present "Cached" links to the pages of Field's site&quot
810.
by Ukv 3y ago
At a glance, the supplementary materials from IBM's paper claim even less: > NorthPole's core array includes 192 MB of flexible memory (768KB of unified memory per core). Assigning 2/3rd of this memory to parameters, such

More ›