dboreham

It's asymmetric warfare. It costs them little to demand a false positive be acted upon, but costs you plenty to refute it.

55 comments and nobody mentioned the incorrect English in the title?

GPUs are already ASICs.

I'd say "some learning" -- I've learned how to operate and maintain an excavator from YouTube without significant struggle. otoh it might be difficult to learn linear algebra that way. But otooh an LLM might be far better than YouTube at teaching mathematics, while being orders of magnitude less costly than a real teacher. Even in the USA proper mathematics teachers are unavailable in many areas.

I mean, I'm concerned about the societal impacts too, but this article is pure nonsense as far as how useful today's AI tools are. I know multiple people with very high paid jobs who have told me they simply couldn't do that job now without AI. I create software much quicker and better than I have in the past 40 years using AI tools. I've learned all kinds of things technical and otherwise from LLMs. This writing is about as useful as the people who said calculators would mean nobody understands mathematics, or that watching tv would turn everyone into a zombie (obviously it was smart phones that did that).

I read the paper's abstract. They're not actually claiming what commenters here think they are. They're saying that there's some sort of pipeline processing for the task of "listening to speech" and when the subject switches focus from listening to one speaker to another, that pipeline has to drain. And the new stream pipeline has to get loaded up. Not actually surprising when you think about how it would have to work. The same thing likely happens with vision.

Not really (except perhaps if it's Stockhausen they're playing). A piece of music played with two hands is still one coherent information stream.

Seems doubtful to me that there will be any hardware in our species specific to processing language. It's going to be the same basic hardware as Chimps. Whatever is going on that's related to us being a social species is all software.

Falklands war shows otherwise.

Different classes of shares are possible, as are other time-varying schemes such as warrants. But that's not what Musk did. The shares all have the same class.

Yes, but all bubbles (except the tulips...) have a real, valuable, new technology at their core. That it's amazing technology doesn't stop the financial side of it being a bubble. In fact it all but ensures it is.

Not the parent but I'm guessing: a) it's expensive and b) you can shoot your feet off.

Sometimes conventional wisdom stops being wise. Also 90% of the people in charge of conventional wisdom have their personal wealth depend on retail investors not selling.

But you can play them if you have 6-finger AI hands.

Not to rain on the parade but my (er...considerable) experience has been that the videos you end up watching to figure out non-trivial songs do not have tabs at all. The teacher generally describes finger positions one finger at a time visually, or will describe fingering as "it's a D chord shifted up to the 7th fret" etc. E.g Uncle Stu -- no tabs, 12-Foot Chain guy has a sketchy tab once in a while, Shutup and Play guy -- no tabs, Jon MacLennan, Justin, Marty (cool Marty) puts up pencil tabs here and there, other Marty (hat guy) does not, etc. I haven't heard Rick Beato play in a long time but I don't remember him putting up tabs either.

Heck I'm going to say it: tabs actually aren't the medium you want anyway. After getting up and running on guitar, players know what all the chords look like so only need the names with perhaps the fret position as a guide to the voicing.

But the tabs would be wrong (since they're invariably wrong in UG).

Hmm. Either I've lost my mind, or you're running a different Docker than me, or you're thinking of some strange scenario such as a Mac where docker is actually inside a VM, or you're wrong.

While there is a feature to do with UID mapping, it doesn't actually work/isn't usable/nobody uses it in current docker iirc.

Therefore root in the container very much is root on the host.

I doubt you need a prescription for most vaccines. Typically there's some sort of guidelines that might say normally given to over 55 years old, but there will be some wiggle room due to other hard to prove preconditions such as family history of xxx or sensitivity to yyy. You say you have one of those things and Costco or the public health dept. will give you the vaccine no questions asked. For example when the COVID-19 vaccine was rolled out it was supposed to be for older people, but also people with various other (not easy to test for) preconditions such as asthma. So everyone who wasn't an anti-vaxxer developed asthma.

Medications are different. Need prescription for that, although $$ and a zoom call can solve that problem.

That said, once built and lit up, it's hard to move a data center to another country.

Serious things tend to be long and tedious and potentially full of proprietary information.

As someone who was there at the time, OSI certainly didn't fail by being "overcomplicated". It failed because a) they charged money to read the standards documents and b) TCP/IP already had so much deployment momentum that nothing was going to supplant it (we see proof of this in the fact that IPv6 also didn't achieve that). Edit: also c) there was no requirement (unlike RFCs) to have an interoperable reference implementation available. So the implementations that were created mostly didn't interoperate.

The waving, and the resulting need to memorize a zillion special cases, put me off Chemistry for life.

I saw the title and assumed an article by Wolfram. But it's by Tim Roughgarden who I know from algorithmic game theory. Anyway, I'll register my membership in the "it's more fundamental than that" camp.

My experience has been that Claude/Opus will ask me questions, although sometimes I have to step in and redirect a bit. Codex/GPT just dives in and cranks code out, which after using Claude for a year I found rather concerning. It wrote working code though. Claude/Fable does something similar (asks far fewer questions than Opus) fwiw.

Ok thanks. Executed on that. I had it build a simple project (actually not so simple since it involved domain knowledge about snowpack and ice) and also gave the same prompt to Opus and to Fable. When I get some spare time I'll write an article highlighting the differences between all three.

Many projects that were done by humans and took a year can certainly not be trusted.

Egads surely Mozilla can produce a blog post that isn't written in AI-speak?

Nudged by this thread, I've decided to switch from Claude to Codex for a bit to see what happens. But...I immediately became lost in their marketing vortex of confusion on plans and pricing. Anyone care to tell me which plan I should be using? On the other side I use the $100 Claude Code plan. We actually have a "Business" ChatGPT subscription already, which seems to be $50/mo/seat. OpenAI's web site offers a set of individual subscriptions (for parity with CC presumably) which I suspect weren't available when we signed up for ChatGPT. I think that in turn happened due to some web site feature it didn't allow for free users (uploading PDFs, something like that). Perhaps I should switch from that business account to an individual subscription for Codex?

But aren't the politicians also corrupt? (or at least most of them) One therefore assumes that any action by congress must be corrupt. This appears borne out by the evidence over the past few decades.

Because it's used in addition to, not in place of, the better RAM.