azornathogron

Is "token" a directly perceivable unit for the LLM? If you ask it "how many tokens are in this sentence?" can it count them (again, not guessing or making a tool call)?

I've never tried it and it might take some thought and effort to conduct an experiment to find out properly, but I would be interested in the answer.

Which are you referring to as the primary comparison?

If I'm understanding the book's structure correctly, the comparison at the end of Part 1 (chapter 18 "Can we live on renewables?") is based on estimating existing demand. Deployment of heat pumps in Britain even today is woefully low, and when the book was written it was no doubt even lower, so of course it wouldn't make sense to include heat pumps as a significant factor when comparing actual demand when the book was written with potential supply.

Chapter 27 "Five energy plans for Britain" discusses large scale ways to make the energy budget "work", and those hypothetical comparisons do include use of heat-pumps. Maybe those numbers are inaccurate and a better hypothetical could be produced today, but I would first put that down to the difficulty of forecasting the future.

It's been a long time since I read the book fully though. Did I miss something?

Heat pumps, and their efficiency (getting more than X Joules of heating for X Joules of electricity) are discussed in chapter 21 Smarter Heating, see for example the diagram and discussion on page 150: https://www.withouthotair.com/c21/page_150.shtml

I imagine the economics (and perhaps technical improvements) of some things may have changed more than the book could forecast. But heat pumps were definitely understood by the author.

Page and Brin together hold over 50% of voting shares, even though their economic share is lower as you suggest. Different share classes.

https://aifundingtracker.com/who-owns-alphabet/

Going to the doctor for expert advice is often inconvenient, or time consuming, or expensive, or stressful. I think a lot of people want, and seek out, information and advice based on their symptoms, as a first step before a possible doctor's visit. Before LLMs, WebMD (and excessive self-diagnosis based on WebMD) was a meme for a while.

So with regard to LLMs, for me the question is not purely "how often does it get it right?" the question is "how does it compare to the sources and self-diagnosis methods people use otherwise?"

Of course, I agree people should be careful with any form of self-diagnosis or LLM-diagnosis.

The way this used to be handled is to sell new versions.

Oh, interesting. That does make sense.

Sounds a bit like imitone https://imitone.com/ (I have no affiliation)

FYI, Imitone has a patent, I don't know what it covers or claims though.

I don't know anything about it, but I find this interesting, perhaps slightly surprising. It doesn't suprise me that it would fade, but, that being a presumably known problem, do the color checkers not have a good cover/case for storage? (something foil backed or similar to really block out the light)? Or do they even fade quickly just from being taken out briefly for use?

Do Zuck or Altman have charisma?

(Jobs I'll give you.)

Neither Mythos nor Gemini are open-weight models (right?) so I'm not sure what model(s) this is actually referring to.

> chronyism (not sure of that spelling)

Since you mentioned you weren't sure: it's "cronyism", no 'h'.

https://www.merriam-webster.com/dictionary/cronyism

This is despite having an etymology that - at least according to etymonline - does come from greek "khronos"

https://www.etymonline.com/word/cronyism

Yes.

From the release notes:

2002-06-17 (2.5.0), "Parse (but do not implement) foreign keys."

At one point there was also a tool which would generate trigger rules to enforce foreign key constraints. (2008 Oct 15 (3.6.4), Added the source code and documentation for the genfkey program for automatically generating triggers to enforce foreign key constraints)

Armadillo Aerospace was Carmack's own rocket company.

And tbh I'm not sure it was ever a plausible contender for commercial success, more like Carmack wanted to play with rockets. But that might be unfair; I would happily accept a correction.

https://en.wikipedia.org/wiki/Armadillo_Aerospace

Surely satellite collision avoidance is not based on visual checks?

I would assume it's a combination of globally shared known trajectories plus radar tracking stations.

They store energy but there's a limit to how quickly you can get that energy out, hence a limit to power. And I suppose you can kinda talk about them "storing power" in the sense that power is what you might be interested in getting out of them.

At least, that's my understanding of it.

What makes it "production grade"? How are you defining that?

Taxis (and Uber etc) also take up space on the road when they only have their driver and no fare paying passenger on board, so I don't see that a Waymo is any worse than that.

Both human-driven and robo-driven taxis are financially incentivised to spend as much time as possible carrying fare paying passengers and as little as possible driving empty to pick someone up.

Anyway, I agree that walking, cycling, and public transit, are all IMHO preferable to any form of taxi.

In type system theory I think what you're looking for is "effect systems".

You make the type system statically encode categories of side-effects, so you can tell from the type of a function whether it is pure computation, or if not what other things it might do. Exactly what categories of side-effect are visible this way depends on the type system; some are more expressive than others.

But it means when you use a hash function you can know that it's, eg, only reading memory you gave it access to and doing some pure computation on it.

You might be referring to a different period, but I'll note that the anti-war protests in early 2003 (immediately before the invasion of Iraq) were quite literally record breaking.

https://web.archive.org/web/20040904214302/http://www.guinne...

It may have in the future. Crubit is one effort in this direction: https://crubit.rs/

For one of my projects my server needs a private key, and it reads this from a file descriptor on startup and then closes the fd. The fd is set up by the systemd unit, which is also configured to restrict filesystem access for the server. So the server reads a key from a file that is never visible in its mount namespace.

I don't doubt you're correct about the incentives, but one point seems amiss...

> If there's any, even the slightest, chance that buying from a business might one day reflect badly on the civil servant in the procurement office, then they won't buy from that business.

You don't think that spending £4.1 million on this garbage might reflect badly on someone?

If you scroll down slightly you get a low contrast button "○ PREFER STILLNESS? READ AS PLAIN TEXT →", which takes you to a plain text version with a rather patronising introduction that says "You chose the quiet version. No animations. No counters ticking up. Just words. That's a valid choice."

Edit: to be just slightly nicer about it: having a plain text version is great, that's a really good thing. But the "that's a valid choice" paragraph is unnecessary and just distracts from your actual article. If I pick the plain-text version it's because I want to just get straight to the point (other people may have other reasons), and I certainly don't need your validation.

Assuming that you are the author of Emissary, this could be a Show HN, I think.

https://news.ycombinator.com/showhn.html

Absolutely.

I think some of these people need to be reminded of the Bill Gates' quote about lines of code:

“Measuring programming progress by lines of code is like measuring aircraft building progress by weight.”

The graph I see in the article displays a single data point per year. You're not going to see seasonal variation in a graph with that resolution. Is there another graph that I missed?

Oh, ok. I misunderstood - I think we agree.

It's not a false choice - "Trust" and "don't trust" are both perfectly viable options. The editor works fine in restricted mode, you just won't have all your extensions enabled.

Yes there was, Asimov added it in Robots and Empire.

"Zeroth Law added" https://en.wikipedia.org/wiki/Three_Laws_of_Robotics#:~:text...