nonfamous

Thanks for the anecdote, but everyone’s advertising experience is different — that’s the whole point of targeted advertising. It doesn’t invalidate the OP’s point, though.

FWIW, I just used the Google app on iOS and got one ad (for artlist.io) before the midjourney.com link. A lot of people use Google this way to get to a named website, btw.

And yet somehow this article fails to mention that with the budget cuts to LIGO one of its two antennas is likely to shut down soon. Without two antennas this research is impossible.

I’m affected by this, and as pissed at Google about it as anyone, but the headline is overblown. The old Nest devices continue to function as thermostats, and the on-device features like scheduling still work. But I need the cloud-based features (particularly remote control via the app), so I went ahead and paid the upgrade tax.

I wonder if that growth comes from grid-scale batteries or from domestic installations. There are huge government incentives for installing a home battery system and connecting it to a VPP, so I wonder if the policy is focused at the domestic or grid level.

It seems like the key metric missing from this report is volume of referrals, reported over time. Ideally, segmented to user-initiated web searches, to filter out things like searches generated via a Spotlight search in iOS.

I’d be very interested to see the trendline of user-initiated search over time.

Coincidentally, and fortunately, I was already in the process of migrating a blog with 10 year’s worth of content from TypePad. It’s not enough.

TypePad’s export process is awful: the output is a poorly formatted .txt file which is hard to parse reliably. The export process itself fails at least half the time. There’s no export process for images or style files. Automated crawl processes fail.

I LOL’ed at this part of the post:

>> If you have any questions, please refer to our Frequently Asked Questions page here.

The “here” link 404s.

Try Microsoft's "Generative AI for Beginners" repo on GitHub. The early chapters in particular give a good grounding of LLM architecture without too many assumptions of background knowledge. The video version of the series is good too.

From the article, the data Tesla withheld from the court revealed:

* There was no record of a “Take Over Immediately” alert, despite approaching a T-intersection with a stationary vehicle in its path.

* Moore found logs showing Tesla systems were capable of issuing such warnings, but did not in this case.

The correct 24h notation would be: Wednesday 01:20.

By the way, if you’re used to 12h time, you can get a similar experience to that described in the article simply by setting your watch/phone to 24h time. After a while your brain just starts to recognize 15:00 and 3pm as the same thing, and there’s no explicit conversion required for you.

According to the paper, the human and AI responses were both delayed by the same amount (depending on message length) to mask the effect of inference speed on the interrogator.

Also it’s not just the app — it’s the fact that the app is running on unsecured private devices in unsecured locations.

This is seemingly unrelated to the Amazon Titan foundational models available on Bedrock. As if things weren’t confusing enough with OpenAI’s naming scheme…

I would expect it to be argued by defendants that since no man (or indeed woman) at Meta actually read the books that were torrented, the First Amendment does not apply here. The question is: does the First Amendment apply to an algorithm?

1. Create AI bot 2. Wait until it gets lots of followers 3. Sell mentions of products by popular AI bot. Profit.

Sure, they faltered at Step 2, but how was this not the plan?

From the summary:

>>> The only hypothesis the committee could develop that provides a plausible but unprovable answer to all these questions and the observed socket failure pattern is that the socket zinc creep was unexpectedly accelerated in the Arecibo Telescope's uniquely powerful electromagnetic radiation environment. The Arecibo Telescope cables were suspended across the beam of "the most powerful radio transmitter on Earth." The other investigations failed to note several failure patterns and provided no plausible explanation for most of them. To answer these questions with empirical evidence instead of only the inferences that can be made from the existing data, a more comprehensive and widespread forensic analysis of "good" and "bad" socket workmanship and the low-current, long-term effect on zinc creep is required.

The Claude terms of service [1] apparently preclude Anthropic or AWS using GitHub user data for training:

GitHub Copilot uses Claude 3.5 Sonnet hosted on Amazon Web Services. When using Claude 3.5 Sonnet, prompts and metadata are sent to Amazon's Bedrock service, which makes the following data commitments: Amazon Bedrock doesn't store or log your prompts and completions. Amazon Bedrock doesn't use your prompts and completions to train any AWS models and doesn't distribute them to third parties.

[1] https://docs.github.com/en/copilot/using-github-copilot/usin...

>>> Systems that are not currently impacted will continue to operate as expected, continue to provide protection, and have no risk of experiencing this event in the future.

Given that this incident has now happened twice in the space of months (first on Linux, then on Windows), and that as stated in this very post the root cause analysis is not yet complete, I find that statement of “NO RISK” very hard to believe.

No restart was needed to cause this crash. As soon as Falcon downloads the updated .sys file ... BOOM.

Are you sure? It looked like a regular NYT ad for the Wordle site. The kind of ad reputable sites *have* to buy now, lest they be gazumped by competitors.

(It does look like the malicious ad is no longer being served, through.)

And how would one know to do that without clicking on the link first and potentially being exposed to malware?

The correct solution is for a public information campaign to inform everyone never to click on any "Ad" link in any Google search ever. It's the only way to prevent our parents and grandparents from being phished, hijacked, or worse.

Good question. The ad looks like an innocuous link to www.nytimes.com, but clicking on the link actually goes to https://www.googleadservices.com/pagead/aclk. (I didn't try actually clicking the link.)

I like to put my name, contact info, and reference link on the last slide. This is when they’ll actually need it (if they want to follow up with you), latecomers won’t miss it, and it’s usually on the screen for a long time during q&a so people have time to jot it down or snap a picture.

It’s not the messages I worry about so much — it’s the contacts.

I can’t use WhatsApp because on iOS it asks for access to your contacts, which I refuse to give. (Facebook already burned me once, there.) So in WhatsApp, all my contacts show up as numeric phone numbers, and there’s no way to change that, so it’s basically useless to me.

Note this is not referring to the Direct File program, the service that allowed some taxpayers in some states to file directly at irs.gov in 2024.

FYI, there's a recipe for running Phi-3 under ONNX on iOS in the linked repository https://github.com/microsoft/Phi-3CookBook/blob/main/md/03.I...

Apple is partly to blame here. One big problem is that there are so many false positives for the "A new iPhone/iPad has access to Facetime/iMessage on your account" warning -- a message that can easily be triggered on all your devices by an automatic upgrade on one of them. If you have an iDevice that is online but not used often, you'll be surprised by this message every few months.

This problem either causes undue concern on behalf of the account holder, or trains them to dismiss these messages without verifying the potential intrusion. Neither are good outcomes for security of the Apple account.

Leam to kem.

This account has been posting simulations of interesting 3-body scenarios for quite a while. It used to be on Twitter but moved to Mastodon. You can check out the archives and play the videos, it’s quite neat: https://botsin.space/@ThreeBodyBot

R itself could be considered a test of this hypothesis, too. It’s been said that elegant, powerful Lisp would be more widely adopted if it wasn’t for all those gosh-darned parenthesis.

Well, at its core R is a Lisp (specifically, Scheme) but with a more traditional syntax (infixed operators, function calls, etc). And it’s fair to say the adoption of R has, indeed, been more widespread than that of Lisp.

Threads isn’t working either.