johnmcd3

* Waymo vehicle entering a dedicated bike lane.

In California, California Vehicle Code § 21209(a)(3) expressly permits a motor vehicle to enter a bicycle lane “to prepare for a turn within a distance of 200 feet from the intersection” -- among other cases. (The vehicle must yield to cyclists in the lane.)

In a way, DeepSeek Prover's subgoal decomposition is a partial-step towards error/proof-state feedback. (DS Prover breaks down a proof into subgoals and attacks each subgoal separately with batched sampling, then puts the pieces back together.)

This is distinct from the approach of the previous SOTA for an open-weights model (Kimina Prover) which generated at the full-proof level.

While it was very impressive to see Kimina's ability to generate medium-length proofs (think AIME-level problems) without sub-goals or feedback at intermediate steps, it's likely that at least subgoal decomposition will be required for longer proofs (think IMO-level problems.)

I certainly agree that where and how error/proof state feedback is best incorporated (training data synthesis / reward function / CoT during inference / etc.) is a fascinating area of research. (It's rumored that GDM's AlphaProof does use proof state / lean feedback already.)

All iPhone 15 models support 4K60 from their USB-C port.

Do you mean just the iPhone 15 Pro models? I understood the base models are USB2 and wouldn't have the bandwidth for 4K60, right?

Had the same experience. The solution is just to use a family member's phone number.

The phone number isn't ever used again and is just used to limit automated signups as far as I can tell.

I believe Elon said they were targeting SN8 for a high altitude flight with 3 raptor engines, hopefully later this year. Various parts are already under construction.

The comparison to early acupuncture studies makes sense and is a good one. Thanks.

My takeaway from your comment is that studies like [14] and [15] are very likely poorly done (like another comment describes early studies of acupuncture).

In any case, I'd expect that in within a few more years, these sorts of conflicting and controversial results should be further disproven.

Thanks.

I've often heard people say that we know that these radio waves are 100% safe because the heating effects of non-ionizing radiation at these doses is not a significant risk.

However, the article itself mentions that "novel EHF [medical] therapies" use only slightly higher (and also "non-ionizing") radio frequencies.

Per the Wikipedia article on EHF therapies, this seemingly similar radiation appears to have studied, proven biological effect: "Low intensity (usually 10 mW/cm2 or less) electromagnetic radiation of extremely high frequency may be used in human medicine for the treatment of diseases. For example, 'A brief, low-intensity MMW exposure can change cell growth and proliferation rates, activity of enzymes, state of cell genetic apparatus, function of excitable membranes and peripheral receptors.'[14] This treatment is particularly associated with the range of 40 – 70 GHz.[15]"

https://en.wikipedia.org/wiki/Extremely_high_frequency#Medic...

Can someone explain to me how we are so confident 5G is safe, if similar, 40-70 GHz radiation at 10mW/cm2 has been shown to "change cell growth and proliferation rates, activity of enzymes, state of cell genetic apparatus, function of excitable membranes and peripheral receptors"?

Why shouldn't there be risk of similar effect at slightly lower frequencies and somewhat lower power?

(To be clear, I believe 5G to be safe with very high probability. But I wish I better knew how to reconcile this information and explain it to others who are much more skeptical.)

There's a big difference between personally trying to fund the US government (which doesn't scale) and his saying that people in or near his position should have to pay higher taxes.

If you doubled the amount of taxes Bill has personally paid over his lifetime, it would amount to a few tenths of a percent of a SINGLE YEAR of US government tax revenue.

I suspect Bill is saying that he-- and other very top earners-- should have paid more.

US Bank took away next day transfers to a bank account via routing/account number (ACH) as they are pushing Zelle more. They also dropped the (previously higher) transaction limits on 3-day routing/acct numb/ACH down to the $5,000 3-day limit on Zelle.

It’s frustrating when they take away a perfectly good service just push their new thing harder.

The mistakes I see made most are:

(1) Not concretely listing the candidate's contribution (vs the team or project description), and

(2) Not focused on the most impressive and relevant items, but instead a laundry list of things unrelated to the opportunity at hand

Some tips and a resume template: https://www.careercup.com/resume

For many applications, making a transcription seems like an unnecessary step and source of errors. Skipping transcription when the user doesn't need it (most cases where I use it) would seem like a way to get some gain, but perhaps at the reduction of debuggability.

Agree. What's really needed is research into (and development supporting) how to combine the expertise from a speech recognition layer with the next layer in a machine learning process. That higher layer contains the domain specific knowledge needed for the problem at hand, and still leverage a speech layer focused on a broad speech data set and speech-specific learning (from Google, Microsoft, the community, etc.)

Today, how richly can information be shared? I see with Google's speech API you can only share a very finite list of domain-specific expected vocabulary.

Why not have speech tools at least output sets of possible translations with associated probabilities? Do any of the top tools allow this?

Then you could at least train your next level models with the knowledge of where ambiguity most exists, and what a couple of options might have been for certain words or phrases...

We agree about it being not JUST old Windows versions being affected.

I replied to your comment because the "old" Windows XP having no patch available was significant here, and I read your comment as saying "old" windows versions were not proportionally more responsible for WannaCry's rapid spread.

Windows XP is still the third largest version of Windows by current installed base (after Windows 10 and Windows 7).

The fact that Windows XP remained unpatched was significant, as there is notable overlap between Windows machines that aren't getting new security updates (at least within a month or two of their release) and Windows machines still running Windows XP.

This vulnerability was, in fact, unusually dangerous, relative to other Windows XP vulnerabilities that have come to light in the last 5 years, and the install base of the "older" Windows XP machines made a big difference in the ransomware's ability to spread.

It's not "old Windows machines," it's XP to Server 2012.

This is incorrect or at least misleading.

Any machine still running Windows XP, is by any reasonable definition, an "old Windows machine." Windows XP was first released in 2001, and actively supported with updates for 12 years. Windows XP hasn't been supported with critical security patches for over 3 years.

Windows Server 2012 is under active support until Oct. 10, 2023, and was patched against this vulnerability in MS17-010. See the middle of the page here: https://technet.microsoft.com/en-us/library/security/ms17-01... If your Windows Server 2012 machine fell victim to this ransomware, it was for the same reason as those running the newer Windows Server 2016 (also vulnerable to WannaCry): because someone didn't apply security patches in a timely manner.

This ransomware was particularly damaging because of it's unusually wormable nature. (Ring 0, commonly enabled networking protocol, no user interaction required.)