Anybody who's looked up recipes online recently knows that they're being used mainly as a way to serve you ads. Before the recipe is a bunch of ad-ridden prose about the recipe itself, with some pictures. Then, at the bottom of the page is the recipe. Google was making the content more easily available, but destroying the ad revenue.

Gotta publish....

They're focusing on this sort of thing:

Betty Smith will give to her mother Gloria Smith (hereinafter "Gloria") one million dollars (the "Payment") on December 1st (the "Payment Date"). Gloria agrees to deposit the Payment at First National Bank (the "Bank") on the Payment Date.

And saying that you only do that in legal writing, not when you're telling a narrative story.

In a narrative story, you might say this:

Betty gave her mom $1M in December 1st. On that date, she deposited it at First National Bank.

There's ambiguity in the narrative story (who did the deposit -- Betty or her Mom?) In legal writing, you're trying to avoid that ambiguity even if you end up with long hard-to-read sentences. If you're telling a story, it's ok to have a little ambiguity and you trust that the reader will figure it out.

It's not surprising that non-lawyers start to use those mid-sentence definitions when they write legal documents -- they're trying to solve the same problem with ambiguity that the lawyers try. It's like saying "novice coders end up using variables just like experienced coders do."

According to the Article, it looks like it was a guy in Houston who got a bit under $291,000.

Not sure how much luck JPMorgan is going to have in the lawsuit department -- people who do this sort of thing aren't exactly the sort of people who are likely to just invest the money. Chances are it's mostly gone.

the fact that the Cyc organisation "failed" to create a general knowledge graph of all human knowledge does not mean that GraphRAG is "not a promising approach". their efforts predate GraphRAG by a _long_ way, and is not related to modern efforts to augment LLMs

i also did not say that this is "easy". indeed i explicitly said that it's "not a magic bullet", and that you need a "high quality knowledge graph" that "requires careful work to prepare and curate".

modern systems using the GraphRAG approach generally are using much smaller knowledge graphs, carefully prepared based on a very specific set of knowledge, such as a support knowledge-base. when done right they produce significantly higher quality results than an LLM alone.

it's still early days for the GraphRAG approach. the first research papers only came out a year ago. it is far too early to decree the approach as "not promising"

An LLM by itself is indeed inherently limited. They are essentially language prediction systems, so their base functionality is to predict what word should come next.

Couple an LLM with RAG (Retrieval Augmented Generation), an approach for combining an external knowledge store with the LLM's ability to generate responses to statements, and you can enhance LLMs quite effectively. Research in this area has been happening over the past 4 or 5 years.

A very promising technique that has emerged this year is GraphRAG, which uses graph databases as the knowledge store.

When you have a suitable graph knowledge-base, your LLM can cite its sources - this greatly reduces problems of hallucination and helps address issues of "accountability".

It's not a magic bullet - you need a high quality knowledge graph as part of your system, and that requires careful work to prepare and curate.

Maybe that's the case where you live but in most of the civilised world a salaried employee works for a restricted number of hours. In Europe (including the UK) that is legally restricted to 48 hours per week. (Overtime is possible, but mandatory overtime is, essentially, illegal.) See the European Working Time Directive.

Note that the online terms are what govern CloudStrike's relationships with small businesses where the size of the business doesn't justify CloudStrike's negotiating a separate agreement. Bigger companies negotiate their own agreements. If you're in the legal department at, say, American Airlines, you don't say "Oh, your extremely one-sided terms of service are perfectly ok with us." Instead, you hand CloudStrike your 40-page agreement that is very one-sided in your favor. And then you spend a couple of months negotiating that agreement.

But, in the final agreement, there will be a section addressing liability. And, in general, CloudStrike is going to have standards of what they will agree to v. what they won't. CloudStrike just isn't going to agree to be responsible for all the harm suffered by its customers -- there's just no way they could buy enough insurance to cover all that risk and, if they did, the price of their product would go through the roof. So, they're going to say something like "Look, you can get your money back (or 3x your money back, or 5x your money back)" or "You can get back at most $5M" and they'll say "If you need more coverage than that, then you need to get your own insurance."

Are you a fossil fuel shill? Because this comment really reads like that, and I'm surprised it's been moderated so highly.

By itself, yes, wind power isn't a great source of electricity. Neither is solar. Both are by their nature unreliable and incapable of providing a sustained supply. That does not make them useless. Instead this is why they should form part of a mix of renewables contributing to an electricity grid.

A grid should have ways of storing excess generation to put that power back into the grid when renewable sources aren't able to contribute. Historically this has been pumped hydro in many cases, but there are other options now available including, but not limited to, battery storage.

Another aspect of having an electricity grid is to send electricity over very long distances, potentially thousands of miles, which reduces the need to store excess power as via the usage of the grid power needs and production can be balanced.

This balancing may not be perfect, and some of the captured power may be wasted, but is that really a problem? Capturing wind is effectively free whereas burning gas has a high cost in terms of the gas input and the carbon dioxide output. (Yes, wind isn't completely free - there's sunk cost for the windmills, and ongoing maintenance costs, but you have similar costs for running fossil fuel plants.). Does it really matter if some of this captured electricity gets thrown away?

Uh. No. Section 230 says that Cox won't be treated as the publisher. That doesn't help it with the copyright issues, which are governed by Section 512 of the Copyright Act (the DMCA). Under Section 512, an ISP isn't liable for the activities of its users if it implements a policy for terminating the accounts of repeat infringers. The claims against Cox were that Cox didn't actually implement that policy, so that liability shield (the "DMCA Safe Harbor") didn't apply to it.

Microsoft ported their BASIC to loads of different CPUs. Commodore BASIC and Applesoft BASIC were both produced by Microsoft. They covered all the popular CPUs back in the day (8080, Z80, 6800, 6809, 6502, etc).

So yes, they started by selling a BASIC that ran on the 8080, but they didn't just stay there.

Interesting side-fact here - Windows NT started out targeting platforms other than the x86 - the x86 version was a port. They did this because they didn't want to be locked into Intel.

(1) GOODs have an obligation to be fit for their intended use. If they're not, then that's either a warranty question or a products liability question. SERVICES, on the other hand, are governed by contract. And, in Google's terms of service for Google maps, they specifically disclaim this sort of liability.

(2) In any case, the suit is in negligence, not products liability. And, there they run into a problem because North Carolina is a contributory negligence state. This means that if the driver could have avoided the accident by acting prudently, then *even if Google breached its duty of care* the driver can't collect.

(3) the "You were told about this multiple times" argument doesn't really go that far. Google can't change something just because it's gotten multiple reports. If it did, then you could easily see a group of teenage pranksters reporting things that just weren't true. And, it's not like they have somebody actually reviewing every image taken with Google Street view.

(4) One thing going against the driver is that it's been 9 years. If it really were the case that a reasonably prudent driver wouldn't have noticed, then you would have expected somebody else to have done the same thing in the previous 9 years.

Ctrl-+A and Ctrl+E work fine for me on the Mac version of VSCode...

I wonder if you've installed some kind of extension that's re-defining those key combos.

Apple Maps works just fine on the web - although Apple themselves don't really seem interested in hosting a site for it, in part because they want to sell iPhones, sure, and probably they'd also rather folks use the app version as it offers slightly better OS integration.

To use Apple Maps on the web just go to Duck Duck Go, search for a place, and then click thru to the maps tab. Many other web sites use Apple's MapkitJS to show maps.

As for preferring Open Street Map, Apple is both a consumer and contributor to the Open Street Map dataset.

> Everyone has a foot

Exactly. And, not only does everybody have a foot, they have a foot with them ALL THE TIME. People use body-based measurements largely because of convenience, not because of accuracy.

Last week, I had to check to see if a washing machine would fit through a doorway. I put one hand on each side of the washing machine, stepped back, holding my hands in place, then stepped up to the door and compared where my hands were with the width of the door. That's a body-based measurement. I could have used a tape measure, but the extra benefit in accuracy wasn't worth the time in finding the tape measure.

They told us the lag in their presentation - it's 18ms