With, I think, a massive grain of salt since this info is unverified and direct from the manufacturer...

Huawei’s official presentation claims their Cloudmatrix 385 supercomputer delivers 300 PFLOPS of computing power, 269 TB/s of network bandwidth, and 1,229 TB/s of total memory bandwidth. It also achieves 55 percent model fitting utilization (MFU) during training workloads and offers 2.8 Tbps of inter-card bandwidth, heavily emphasizing its strength in networking.

| Spec            | NVL72 (Nvidia) | CloudMatrix 384 (Huawei) | Better? (%) |
|-----------------|----------------|--------------------------|------------|
| Total compute   | 180 Pflops     | 300 Pflops               | 67%        |
| Total network bw| 130 TB/s       | 269 TB/s                 | 107%       |
| Total mem bw    | 576 TB/s       | 1,229 TB/s               | 113%       |

A chart titled "What Kind of Data Do AI Chatbots Collect?" lists and compares seven AI chatbots—Gemini, Claude, CoPilot, Deepseek, ChatGPT, Perplexity, and Grok—based on the types and number of data points they collect as of February 2025. The categories of data include: Contact Info, Location, Contacts, User Content, History, Identifiers, Diagnostics, Usage Data, Purchases, Other Data.

• Gemini: Collects all 10 data types; highest total at 22 data points
• Claude: Collects 7 types; 13 data points
• CoPilot: Collects 7 types; 12 data points
• Deepseek: Collects 6 types; 11 data points
• ChatGPT: Collects 6 types; 10 data points
• Perplexity: Collects 6 types; 10 data points
• Grok: Collects 4 types; 7 data points

I was reading this article about the NYT's suit against OpenAI. OpenAI argued that NYT couldn't sue for damages because it had been "too long" since the infringing started, and since NYT "must have known" that OpenAI was doing it, they lost the privilege of collecting damages (IANAL but I think it's because the Doctrine of Laches). In any event, the judge sensibly threw this argument out, telling OpenAI they hadn't demonstrated that NYT could have known the size or scale or timing of the any alleged infringement.

This made me think: now that the cat is out of the bag and everyone DOES know that everything on the Internet (and beyond) is being fed into AI factories, do we as creators have an obligation to somehow collectively sue LLM makers so that laches can't be used as a defense in the future?

While not the gigantic uber-canines of fantasy lore, these pups will become roughly-gray-wolf-sized dire wolves, and represent the first de-extincted animal species, raising a number of ethical questions about returning animals to ecosystems that may not be stable for long.

The foundation will focus on improving ActivityPub and the user experience, informing policymakers, and educating people about the fediverse and how they can participate. They currently have some backing from Meta, Flipboard, Ghost, Mastodon, and others, and the Ford Foundation has also offered the organization a large grant to get the project started. In total, SWF is closing in on $1 million in financial support (or was, as of September) (from TechCrunch)

A domestic breeding program kept these birds from going extinct. An initial reintroduction to their native habitat on the big island was halted after their natural predators proved too adept (or the coddled crows proved not adept enough, I guess). So they're now being relocated to Maui.

File this under "small wins". I had been banging my head against a technical problem for most of the day yesterday. As I slipped into bed around midnight, I suddenly knew the solution. Despite the call of the pillows, I dragged myself out of bed, down to the laptop and took a full 30 seconds to write it down -- and thank goodness, because by this morning I had forgotten about it again!

Crowds and water have more in common than you'd think - they both flow like a fluid, with predictable patterns that can turn perilous if not properly managed. Looks like the physics of human herds is no bull, as researchers have uncovered the fluid dynamics behind dangerous crowd crushes.

Using Reddit's popular ChangeMyView community as a source of baseline data, OpenAI had previously found that 2022's ChatGPT-3.5 was significantly less persuasive than random humans, ranking in just the 38th percentile on this measure. But that performance jumped to the 77th percentile with September's release of the o1-mini reasoning model and up to percentiles in the high 80s for the full-fledged o1 model.

So are you smarter than a Redditor?

When even Cory Doctrow starts to sound like an optimist I have to give myself a reality check as it usually means I'm heading off the deep end. But in this case it just rubs me the wrong way that he talks about Mastodon and Bluesky in the same breath -- one is not like the other.

Originality.AI looked at 8,885 long Facebook posts made over the past six years.

Key Findings

• 41.18% of current Facebook long-form posts are Likely AI, as of November 2024.
• Between 2023 and November 2024, the average percentage of monthly AI posts on Facebook was 24.05%.
• This reflects a 4.3x increase in monthly AI Facebook content since the launch of ChatGPT. In comparison, the monthly average was 5.34% from 2018 to 2022.

Yet another entry from the truth-is-stranger-than-fiction department, as drug-addicted rats have turned Houston’s police evidence storage into their personal stash house.