与「Couple」相关的搜索结果

I had @spacexai Grok 4.6 follow Fable's plan, and with just a couple of nudges, it was able to repeat this feat in 1h 24m using 8.6M tokens at a cost of ~$55 at per-token pricing. That's about 1/10 the cost of the Fable implementation for the same work!

I used Grok 4.6 to build a 2D agent simulation of The Office. Here are notes and observations on 4.6: It is a huge step up from 4.5 - both in the quality of visual outputs and in the breadth of work: personality, features, and finding usable assets for this project. The editorial content and writing were good - I like the "feel" of this model. One of the biggest wins for 4.6 is speed. The throughput (~80 tok/s) and latency are so good that it feels Composer-adjacent - something @ericzakariasson called out in his writeup. That speed + cost + performance makes for a true "workhorse" model that's practical for both personal and work use. My goal was to simulate The Office through "agents" that behaved like the characters - including all of the neuroticism and idiosyncrasies that made the show so funny. I knew what I wanted artistically - I've built similar 2D apps with pixel art. I asked Grok 4.6 to research psychological modeling, pull character descriptions, find references for the cast, and assemble 2D sprites with the Universal LPC sprite sheet generator and some RPG tilesets I grabbed online. I rattled a brief into the @cursor_ai mobile app late one night at the Cursor office and woke up to an MVP. Then I spent a day or so spamming cloud agents until the simulation worked how I liked. The cool part is that Grok did the simulation, art, and research on its own: - Every couple of minutes, each person scores a few possible actions and picks one. They pathfind around desks, work, grab coffee, talk at the cooler, hide in the annex, or end up in a meeting. - Click someone and you can inspect traits, needs, mood, relationships, and whatever they're doing right now. You can also speed time up, force Michael to call a meeting, or roll a random office event. - Each character has Big Five scores that bias what they tend to do. Needs like work, social, caffeine, status, and comfort rise over time and fall when they do something about them. - Grok reconstructed the layout from Dunderpedia (Office Wiki) and online sources - Michael's office and the conference room on the north glass, Pam at reception, sales and accounting in the bullpen, the annex to the east, restrooms and the warehouse down the stairs. - Grok helped me find a few sprite sheet packs to implement. I was surprised at how accurate some of the character sprites were - from hair color to clothing. The combination of a performant model, cloud workspaces, and high throughput made for a really fun devex - the amazing folks at @cursor_ai and @SpaceXAI make my job easier everyday. A huge shout-out to @_Brian_Zhang, whose Notion hackathon project inspired this exploration.

Claude's watermark probably doesn't work how you think. As the CTO of GPTZero, I'll explain how Anthropic, Google and OpenAI are building text watermarking in this brief explainer and whether it can be defeated. Almost all forms of watermarking that are fast and cheap enough for a frontier lab have the same formula, following the KGW method: In generation: 1. Let's say you've generated n tokens so far. Take those n tokens + a secret key to generate a random hash 2. Use that hash to randomly reweight the probabilities for the n+1 token, and then sample from that new distribution. In the simple case, you could split 50% of all English words into a green or red set based on your hash, and boost the probability of words in the green set. For watermark detection: 1. For each token, see if it was in the green or red set. 2. To do this, recreate the hash based on the secret key and the text preceding the current token. Then, recreate the green and red set of words. 3. Once you've checked all the words in the text, if the next token is selected disproportionally from the green set more than 50% of the time, you claim the text has the watermark. I can tell you want to ask the following: 1) Isn't it easy to mess up the hash if you paraphrase the text? The answer is mostly yes, however, you can use a statistical model to get your hash instead of a deterministic function (SIR, Adaptive Watermark). Since the entire watermark is probabilistic, this is fine. 2) Doesn't this make the text much worse? The answer is yes, it does - Yes, it does – but for most people, it's imperceptible (Google claims in human feedback study with 20,000 texts), since there are exponentially many ways to write the same paragraph. DiPmark does something more sophisticated to avoid shifting the text distribution on average. Of course, watermarks fail on short text or highly predictable texts like "2+2=4". 3) Shouldn't it be easy to figure out the green and red sets? The answer is no. You would need an exponentially large number of samples from the watermarker to reconstruct those sets exactly, but it's a risk if the detector is open to the wild (Watermark Stealing) Still, there are couple challenges that a frontier lab needs to overcome: 1. Their watermark needs to work token-by-token because they are streaming their text to users. Many watermark methods plan sentences or paragraphs at a time, or change the text after its entirely written, in order to make their watermark robust to paraphrasers, and a frontier lab cannot afford to do this yet (SemStamp, PostMark) 2. If the secret key leaks, the watermark is busted. To avoid a large blast damage from this, you need to have a couple secret keys in rotation. 3. There are some texts, like code, that cannot be arbitrarily changed, otherwise the code will break. In those cases, the watermark needs to selectively change words in parts of the text that can tolerate synonyms (i.e. like variable naming) - see SWEET, EWD, Invisible Entropy. 4. They will need to educate their users on how to deal with false positives and false negatives of a detector, which is a big challenge (one we put a lot of effort into) So, how do I see this playing out in the next 6 months? 1. If Anthropic releases the watermark detector publically, I think they defeat their own watermark. People find reliable watermark removal strategies by testing against Anthropic (AI detectors like GPTZero have an advantage here because they can train against these adversaries once they become popular). 2. If they keep the detector private to the government, like Google has done, it's "safer". However, there are some papers showing trained approaches that work robustly to zero-shot break watermarks without any data, simply because they try to write the text just like a human (Zhang et al. 2024, Watermarks in the Sand). Also, making your detector makes it battle-tested and stronger long-term (my experience). 3. In my testing, the watermarks don't survive intense paraphrasing (especially if you combine word choice and syntax attacks), or human text substitution (rewrite your AI text by plagiarizing human authors). The free paraphrasers I've tried have quickly bypassed Google Deepmind's SynthId for what it's worth. 4. All-in-all, frontier labs are likely okay with this because they expect most users to not attack the watermark, and also because they + European regulators likely don't care past a certain point - its good enough. 5. Overall, I think users of frontier LLMs will not really care about this, because 1) they don't realize watermarks are there, 2) EU will force everyone to conform, 3) this seems more like regulatory hoop-jumping than an earnest effort from frontier labs to expose LLM use Lastly, people's first concern shouldn't be watermarking, it should be AI detectors! If you're posting, "its not X, its Y!!", I don't think the watermark is going to make a difference :)

Andrew Ross Sorkin today on his @Starlink experience on United Airlines: "Amazing. I wonder what it's going to mean for other airlines . If you were traveling from NY to LA for the next couple years, if you had a choice between Starlink and no Starlink, you'd choose Starlink. I think this is going to be huge for them."

This story serves as a perfect exemplar of how parasitic ideas coupled with suicidal empathy have fully gripped academia. Academic honesty takes a backseat to the plagiarist's skin color. Truth takes a backseat to promoting voices of color. Integrity takes a backseat to celebrating "marginalized" voices. It is a grotesque inversion of a well functioning moral compass.

"You need bats in today's game if you're going to compete." @JeffPassan breaks down some of the moves that have been made the last couple of days after the Phillies acquired Luis Arraez⚾️