meatmanek
Don't get me wrong, I understand why it makes sense for a company to have the ability to block spammers, scammers, and malware without proving a case in a court of law, but I feel like there's a spectrum of importance, where the more potentially harmful the action is to the user, the more process is due to the impacted user to make sure the action is justified. Examples from least- to most-important:
- removing individual content that has little real-world importance: comments, social media posts, etc.
- banning accounts on platforms that generally have low real-world value (other than sentimental value) to the user: reddit, HN, non-monetized Youtube accounts, etc
- banning accounts on free platforms that have a very high cost to the user of being banned: Gmail (keys to the user's digital kingdom, the rest of the Google account), Facebook (real-name policy means that it's difficult to start over), high-notoriety accounts on other social media, etc.
- removing individual listings/etc on platforms where the user makes money: eBay, monetized Youtube videos, paid App/Play store apps, etc
- Banning accounts on platforms where the user makes money and the account itself has some intrinsic value (reputation, following, etc): eBay, monetized Youtube accounts, App store accounts, social media accounts for businesses, etc.
- Locking up or taking the user's money: locking the user out of a bank/brokerage/crypto/payment platform account; rescinding money (or other things with monetary value): bank account balances, payment platform balances, unpaid Youtube/other monetization earnings, loyalty/credit card points, etc.
- Banning accounts that are essential for the user to participate in society: bank account, cell phone plan, identity providers, etc.
Classic XOR encryption is like this. You can make a given ciphertext decrypt to anything you want by XORing the ciphertext with the desired plaintext to get the key.
Therefore, just because you've found a way to decode something to plausible-looking text doesn't mean you've found the correct key.
The "most attractive quadrant" is completely meaningless. The whole point of a Pareto curve is that each point on the curve is better than everything else on at least one dimension, and that you can make these comparisons without placing a value judgement on the relative importance of the different metrics. If you make a composite score of the two metrics (any monotonically non-decreasing function, e.g. a weighted sum with non-negative weights), that score will always be maximized by one of the points on the Pareto frontier.
So going by the numbers in the 2nd chart (1st AA chart) from TFA alone:
- there's no reason one would choose Deepseek V4 Pro 0813 (max) even though it's in the "most attractive quadrant", because GLM-5.3-Flash is both cheaper and scores better.
- Claude Fable 5.1 (max with fallback) on the top right* could be your most attractive option if you need the best scoring model and don't care about cost, even though it isn't in the "most attractive quadrant"
- The un-shown model off the left side of the chart could be your most attractive option if you just need lots of cheap tokens and don't care about quality.
* I also dislike the way they place the labels, and that grey line connecting the label to the point is way too subtle.
By 10x you mean 10% more, and by ozone layer you mean the fact that the southern hemisphere's summer coincides with Earth's perihelion while the northern hemisphere's summer coincides with aphelion.
https://www.abc.net.au/news/science/2025-02-04/sun-summer-uv...
The Huggingface link on https://narilabs.com/product/stt/ links to https://huggingface.co/Qwen/Qwen3-ASR-1.7B , not anything under https://huggingface.co/nari-labs
Is the ASR inference engine open source as well?
EDIT: I see I have $5 of "pre-paid API credit". Is this a one-time thing or is that included with my plan?
The processing pipeline also started imposing its own opinions on your photos, adjusting the brightness of dark areas and bright areas to even things out, increasing saturation, etc. This has generally trended towards more processing over the years.
In a picture from a standalone camera (or an old phone back when the processing pipeline was basically just debayering, white balance, chroma denoise, and a gamma curve), you'll actually have contrast between shadows and highlights, which, when done well, looks great. When done poorly, it looks terrible (blown highlights, noisy shadows), which is the whole point of the processing pipeline -- on average (random user, random scene), you're better off reducing the overall contrast in the image with local adjustments, even if that turns some excellent photos into boring ones, because it'll turn a lot more bad photos into acceptable ones.
This is similar to the conclusion from one of the blancolirio videos about this crash[1]:
> The operation that dispatched this flight should have never dispatched this flight in the first place to begin with. There are simply too many variables working against you to accept this flight. And in my opinion, this is another classic situation where we're losing a lot of young pilots just trying to build up flying time to that magic 1500 hours to get on to a better job and a better future with the airlines. In my opinion, we should have never implemented the 1500 hour rule. These kids would have been much better off if they had been hired by the airlines at 500 hours and been trained and kept in a safer, more controlled environment in the right seat of the airline that they're going to work for for the rest of their career while they build up their time there to get to 1500 hours, where they can then get their ATP rating and then upgrade to the left seat; instead of sending all of our young pilots out there to do all kinds of risky flying operations in an effort to build up the magical 1500 hours as required by a knee-jerk reaction by Congress to assuage the family members of the Colgan Air crash victims. Get rid of the 1500-hour rule.
> Kingett has Cerebral Palsy, CP, in addition to his blindness. He’s had both ever since he was a baby. He has Retinopathy of prematurity, commonly referred to as ROP. Even though he is ambulatory, his cerebral palsy makes it difficult to walk long distances or to stand for lengthy periods of time, which is probably why writing fell right into his lap!
So it sounds* like he uses his cane from his wheelchair and/or when walking.
*unintentional metaphor
2. It's not uncommon for people to use terms like "watch", "see", "hear", "listen", etc. colloquially / metaphorically; e.g. "I heard that X did Y" by someone who, pedantically speaking, _read_ that X did Y; someone typing "listen to me" on a chat app when they really mean "please read my words carefully and internalize them"; etc. If I were playing a video on on my phone, listening to what is being said but not actually watching the screen, I'd probably still say I "watched" the video.
Apple: https://support.apple.com/en-us/102515
> If Location Services is on, your device will periodically send the geo-tagged locations of nearby Wi-Fi hotspots and cell towers to Apple to augment Apple's crowd-sourced database of Wi-Fi hotspot and cell tower locations.
Google: https://support.google.com/android/answer/15157297?sjid=1648...
> When Location Accuracy is on, Google periodically collects information about the locations of wireless signals and sensors observed by your device to crowdsource location estimates. This helps everyone find locations better.
Mozilla used to run a very similar service: https://en.wikipedia.org/wiki/Mozilla_Location_Service
Not to mention truly crowd-sourced databases like wigle.net.
Even if you are upgrading every single year, it works out roughly the same to just buy every phone outright and then trade in last year's model -- you typically get ~half your purchase price back as trade-in value after a year anyway.
https://huggingface.co/prism-ml/Bonsai-27B-gguf#weight-repre...
> Weight Representation: Q1_0_g128 Each weight is a single sign bit: 0 maps to −scale, 1 maps to +scale. Every group of 128 weights shares one FP16 scale factor. > Effective bits per weight: 1.125 (1 sign bit + 16-bit scale amortized over 128 weights)
The Bonsai ternary model also uses this group-wise scaling, so the theoretical bits per parameter is 1.71 instead of ~1.58; additionally, they actually store trits as 2 bits; with the group scaling factor it becomes 2.125 bits/weight: https://huggingface.co/prism-ml/Ternary-Bonsai-27B-gguf#memo...
I guess this is assuming some scanning rig (camera scanning?) where it only takes a few seconds per photo. With a flatbed scanner, scanning the backs is almost certainly going to take longer than just typing dates in manually.
Phase 1:
- Artist lists tickets as "$20"
- SH/TM charges $20 + $15 convenience fee
- SH/TM pays the artist $30 and keeps $5
- Artist gets to look like they're kind/generous for pricing so affordably while making some of SH/TM's cut
- SH/TM hold back some percentage of tickets (possibly most or all of them) to sell on the "resale" market
- Those "resale" tickets sell near the market-clearing price, either earning higher margins than if they had sold at list price, or ensuring that most of the seats in the venue are filled.
- SH/TM kick back some percentage of revenue from resale tickets to the artist. - having a log of the most prominent J-space tokens during your customer support chatbot's interactions with a user, so you can have more introspection into why a particular outcome happened
- being able to detect certain thoughts associated with undesirable behavior (hallucinations, overstepping authority, lying, etc.) and trigger some sort of remediation (e.g. upgrading to a better model, redirecting to a human, forcing tool calls)That's one of several reasons why a citation is warranted.