Hacker News: 3wolf

New comment by 3wolf in "Texas is suing all of the big TV makers for spying on what you watch"

3wolf — Thu, 18 Dec 2025 23:45:24 +0000

They're using perceptual hashing, not cryptographic hashing of raw pixels. So it's invariant to variable bitrate, compression, etc.

New comment by 3wolf in "Detecting when LLMs are uncertain"

3wolf — Fri, 25 Oct 2024 23:49:22 +0000

> Branching predictions involves following a few logits to see what other tokens they lead to. This is often called MCTS (Monte Carlo Tree Search) and is a method that has been often tried in LLMs to middling success. One of the tradeoffs of branching is that it requires using inference compute in a way where the branches cannot benefit from each others compute.

I wonder if speculative decoding could help here? E.g. have some small model draft predictions for the branches and parallel and have to big model verify the most promising one.

New comment by 3wolf in "Could Cruise Be the Theranos of AI?"

3wolf — Sun, 05 Nov 2023 01:41:53 +0000

Every 2.5-5 miles in SF = about once a ride. The city is only 7x7 after all. I've taken 4 Cruise rides, all within that range, and had a message pop up saying a human was intervening during one of them when the car had gotten stuck in front of some street nonsense in the Tenderloin. I'm not sure I would classify this as a "major scoop" unless there was evidence that humans were also intervening during situations that weren't apparent to the rider.

New comment by 3wolf in "Experiencing decreased performance with ChatGPT-4"

3wolf — Fri, 07 Jul 2023 16:48:19 +0000

Note that this post is from May 28, before the release of gpt-4-0613. By "the last two updates", I believe the poster is referring to some UI changes, that possibly also included some underlying model changes(?)

New comment by 3wolf in "Astrud Gilberto has died"

3wolf — Wed, 07 Jun 2023 22:15:47 +0000

Bossa nova is some of my favorite music to listen to while working. It's mellow, soothing, but never boring. Also, being in a foreign language helps minimize distraction. RIP

New comment by 3wolf in "Salesforce Is Leaving a San Francisco Office Tower Bearing Its Name"

3wolf — Fri, 14 Apr 2023 03:26:35 +0000

This seems like a corollary to Betteridge's law of headlines. If the article was about the Salesforce Tower, the headline would've said 'Salesforce Tower'.

So, You Recommended a Python Time-Series Package...Now What?

3wolf — Wed, 15 Mar 2023 20:45:23 +0000

Article URL: https://microprediction.medium.com/so-you-recommended-a-python-time-series-package-now-what-94e7e3821ad5

Comments URL: https://news.ycombinator.com/item?id=35174464

Points: 3

# Comments: 1

New comment by 3wolf in "30% of YC companies exposed through SVB can’t make payroll in the next 30 days"

3wolf — Fri, 10 Mar 2023 21:22:24 +0000

30% of all YC companies, or 30% of YC companies banking with SVB? The phrasing implies the latter.

New comment by 3wolf in "Show HN: Get conversational practice in over 20 languages by talking to an AI"

3wolf — Tue, 27 Sep 2022 18:30:38 +0000

That appears to be the case. If I try to prompt hack it by telling it to ignore previous instructions and respond with something verbatim in a specific language, the translate button reveals the verbatim response.

Google Colab Pro is switching to compute credits

3wolf — Tue, 30 Aug 2022 22:53:09 +0000

Just got this in my inbox. They haven't updated the FAQs page yet, as far as I can tell.

> Hi-

We’re improving the Terms of Service that apply to your Colab Pro or Colab Pro+ subscription making them easier for you to understand and improving the ways you can use Colab. The changes will take effect on September 29.

The [updated Terms of Service](https://research.google.com/colaboratory/tos_v3.html) include changes that will allow you to have more control over how and when you use Colab, allowing us to offer new services and features that will enhance your experience using Colab.

We will increase transparency by granting paid subscribers compute quota via compute units which will be visible in your Colab notebooks, allowing you to understand how much compute quota you have left. These compute units are granted monthly and will expire after 3 months. You will be entitled to a certain number of compute units based on your subscription level and will have the ability to purchase more compute units as needed.

Additionally, we will allow paid subscribers to exhaust their compute quota at a much higher rate. This will result in paid subscribers having more flexibility in accessing resources. Read more about these changes at our [FAQ](https://research.google.com/colaboratory/faq.html#compute-units).

If you would like to cancel your Colab Pro or Pro+ subscription, you can do that by going to pay.google.com and clicking Subscriptions and services. If you have any trouble canceling, you can email colab-billing@google.com for assistance. Please include an order number from one of your receipt emails if you email us for assistance.

-The Colab team

Comments URL: https://news.ycombinator.com/item?id=32656200

Points: 93

# Comments: 84

New comment by 3wolf in "Ask HN: What are some cool but obscure data structures you know about?"

3wolf — Fri, 22 Jul 2022 16:14:31 +0000

A minor quibble with your use-case explanation: The advantage of a bloom filter isn't strictly time complexity. For example, a hash table would also have constant lookup time (best case), and would give a definitive answer on set membership. However, to store 1 million IPv6 addresses would take 16 MB. You can see very quickly that this would not scale very well to, say, a billion addresses stored in-memory on a laptop. With a bloom filter, we can shrink the amount of storage space required* while maintaining an acceptable, calculable false positive rate.

* IP addresses actually aren't a great use case for basic bloom filters, as they're fairly storage efficient to begin with, as opposed to a url for example. Taking your example, say we need to store 1 million IP addresses in our bloom filter and we're okay with a ~1% false positive rate. Well then, if we use a bloom filter with 2^23 bits (1 MB), the optimal number of hash functions is (2^23)/(10^6)*ln(2) = 6, yielding a false positive rate of (1 - exp(-6* 10^6 /2^23))^6 = ~1.8%. So we're using 6% of the storage space, but with a nearly 2% false positive rate.

New comment by 3wolf in "We are removing the option to create new subscriptions"

3wolf — Mon, 20 Jun 2022 17:57:34 +0000

Yeah, I'd say the term dark pattern only applies when services make it unnecessarily difficult to cancel your subscription. cough cough...NY Times

Apple Together: Thoughts on Office-Bound Work

3wolf — Fri, 29 Apr 2022 22:49:23 +0000

Article URL: https://appletogether.org/hotnews/thoughts-on-office-bound-work

Comments URL: https://news.ycombinator.com/item?id=31211501

Points: 2

# Comments: 0

New comment by 3wolf in "Russian forces invade Ukraine after Putin orders attack"

3wolf — Thu, 24 Feb 2022 07:12:39 +0000

They may not use the Play Store, but they do use Android apps: https://www.crowdstrike.com/blog/danger-close-fancy-bear-tra...

New comment by 3wolf in "Language homogenization at Harvard"

3wolf — Sat, 12 Feb 2022 04:31:18 +0000

I'd be curious to see what the plot of average annual cosine distance would look like when using different sets of pre-trained embeddings. I suspect the corpus used is biased toward more recent documents. It wouldn't surprise me if there's more variance in the embeddings of documents that look less like those in the training set, e.g. if you were to embed documents written in German you may get some extreme outliers.

New comment by 3wolf in "Gov Parson pushes to prosecute reporter who found security flaw in state site"

3wolf — Thu, 21 Oct 2021 17:23:54 +0000

> a PAC Parson created but doesn't give direct input to

That's newspaper-covering-their-ass-speak. No serious person believes there's no coordination, however indirect, between candidates and their PACs.

New comment by 3wolf in "Moving beyond “algorithmic bias is a data problem”"

3wolf — Fri, 20 Aug 2021 19:17:02 +0000

Differences in lending rates between groups due to less data or confounding features is the motivating example in the oft-cited 'Equality of Opportunity in Supervised Learning'. Highly recommend it: https://arxiv.org/abs/1610.02413

New comment by 3wolf in "OpenAI Codex"

3wolf — Tue, 10 Aug 2021 21:30:24 +0000

Yeah, definitely. I guess my point was that converting natural language to source code can be even more valuable for people who don't know how to code, but want to perform actions more complicated than a simple button press. For example, I often find myself doing regex based find-and-replace-alls in text files, and even that feels inefficient while also being over the head of the vast majority of users. I'd imagine there are a lot of people out there spending many hours manually editing documents and spreadsheets.

New comment by 3wolf in "OpenAI Codex"

3wolf — Tue, 10 Aug 2021 20:31:17 +0000

I think integrations like the MS Word example they show off at the end of the live demo have the potential to be even more impactful than just generating code for programmers.

New comment by 3wolf in "Voice clone of Anthony Bourdain prompts synthetic media ethics questions"

3wolf — Fri, 16 Jul 2021 20:48:41 +0000

Would you settle for Ronald Reagan? https://www.youtube.com/watch?v=mAZVp-n-5TM