Hacker Newsnew | past | comments | ask | show | jobs | submit | dpe82's commentslogin

The slop generator's gotchu: https://suno.com/s/gT2YhaQyNdlkzgeU

NVidia has enormous operating margins, so a competitive solution doesn't have to match or beat NVidia's scale efficiencies; it just has to beat delivered cost.

One objective of the project might be simply to provide credible negotiating leverage when dealing with existing suppliers like NVidia. You don't have to deploy at scale for that to work, but you do have to look like you could if pushed hard enough.


> NVidia has enormous operating margins, so a competitive solution doesn't have to match or beat NVidia's scale efficiencies; it just has to beat delivered cost.

But then that means you have no actual moat against the behemot, right? Your competitor can move into the market as soon as they want to, at much better cost (so at slightly better price)... and Nvidia certainly can adapt much faster around hard hardware specs innovation than a new entrant ever could.


But they are also PURCHASING from nvidia so any time nvidia lowers their prices they save money.


Those are not OpenAI's concerns - they just need to scare NVidia enough to lower their prices more than they'd otherwise want.


> There are simply too few historians trained to read them to be able to cope with the supply.

Surely this is something AI tools will (eventually?) be able to help with?


Posted yesterday (2023 article): https://news.ycombinator.com/item?id=49282415


Yes AI can help, especially in perhaps the "easy" OCR. but where that hard bit comes is where stuff is missing, or the signal is really rather low.

Its also new, or likely new to human knowledge which means that LLMs are not much help without strong guiding.

_if_ I retire, I would love to be translating this stuff until I loose my marbles.


I currently use AI to help me decipher software that was written two years ago.


Another thing to add to all this is FAA has been steadily removing from service non-GPS navigation aids.


Right? "The government" as a whole can't have it both ways. Eliminate older navigation systems because GPS is everywhere, AND run around jamming GPS.


Yes, but…

They are enhancing the navaids that remain. Greater range, and a documented network of airports serviced by ground based instrument approach procedures.

The shrinkage of total number of navaids concerns me from a rural support perspective but less so as a pilot. What does concern me as a pilot is even the MON navaids are having maintenance issues resulting in outages.


> even the MON navaids are having maintenance issues resulting in outages

Which is exactly why we previously had such a dense network: to provide redundancy in the face of inevitable outages.


Not sure anything is enhancing. Approaches have… always been documented? What has greater range - VORs? I haven’t seen any. I think it’s reducing everything to MON that’s it.


MON adds four new SSVs: VL and VH for VORs; DL and DH for DME. There's range changes in VOR, DME, and VORTAC.

See AIM 1-1-8. NAVAID Service Volumes https://www.faa.gov/air_traffic/publications/atpubs/aim_html...

"documented network of airports" is a poorly worded reference to MON airports as distinguished from MON navaids.


`claude update`


Humans (exhibiting "general intelligence") are tool builders; virtually all our capabilities stem from our ability to create and use tools - often extremely specialized to a task. Why would an artificial general intelligence be any different?


Seems roughly in line with UCSF's entry-level post-doc compensation: https://postdocs.ucsf.edu/pay-and-benefits

That too, of course, is too low to live comfortably on your own in SF.


While single and working at a startup back in the dot-com days, I shared an apartment in the Richmond district with two other people. It didn't seem all that unusual. I assume people still do that?

Seems like it would be okay temporarily until you get a raise and the stock options kick in.


> Seems like it would be okay temporarily until you get a raise and the stock options kick in.

In practice it doesn't work like that, usually people who start with those kind of salaries get stuck as a permanent underclass in the Bay. When I was working in biotech out in SF plenty of people were getting like 45K and just expected to make it work. The places who were really on top of it would be pretty overt about making examples of people regularly to remind the staff that another guy or girl from India could be imported in tomorrow.

> While single and working at a startup back in the dot-com days

Unfortunately you don't have a good perspective, since you were sort of carried along and "made it" by dint of timing and location.


These days when you accept salaries that low you’re effectively locking yourself into them for a good long time.


It's illegal in California for a prospective employer to ask about your previous salary.


Hence the existence of services like The Work Number.


Once I learned this exists, I blocked it from ever being requested. And I check my payroll settings to never report income.


*at your current job


"Comfortably" varies by person. I know many people living fine on less in San Francisco.


> *English Common Law* has entered the chat


Still romanian. "The common law—so named because it was common to all the king's courts across England—originated in the practices of the courts of the English kings in the centuries following the Norman Conquest in 1066" [1]

France (and Normandy) was conquered by Rome before it was granted to Rollo, a Viking. A few generations later, his descendant William the Conqueror claimed the English crown. Just like what happened with Gauls and Romans, the conquered people adopted large parts of the language and customs of the conqueror. That's why the English language is ~25% of Roman origin for it's vocabulary. I suppose it is the same for English common law.

[1] https://en.wikipedia.org/wiki/Common_law#History


After 2000 years everything mixes together at least somewhat so trying to draw hard lines is an exercise in pointless semantics, but it's worth acknowledging that as a system, common law is pretty distinct from the Roman tradition of civil law: https://en.wikipedia.org/wiki/Civil_law_(legal_system)


This brings to mind the wonderful Econtalk episode about Bruno Leoni [0]. The beginning of the podcast describes his untimely passing, which almost seems a Cohen brothers movie plot.

  So, we pore over Supreme Court cases on the First Amendment, for example, to 
  try to interpret what tests we will use to determine whether something is 
  going to be unconstitutional law. Leoni didn't want that. He argued that--and 
  again, he was proud of the Roman law contribution. He said that the Roman 
  jurist was a sort of scientist: that the object of his research was a 
  solution to cases that citizens submitted to him for study. So, an 
  industrialist or a scientist might look to a physicist to engineer a 
  technical problem. So, private Roman law was something to be described or 
  discovered, not something to be enacted. So, over time, these principles 
  emerge.
0. https://www.econtalk.org/the-underrated-bruno-leoni-with-mic...


Interesting. As an unhappy user of the court system I have made these same points. Although I linked it back to the enlightenment.

The law should be akin to the scientific method. Unfortunately like science it is full of humans, so reputation is important, etc, etc. unfortunately the judiciary is far worse than the scientific community.


I don't understand why you're getting downvoted.

I've used GPT-5.5 and Opus both for FPGA design with good results. We built a lot of tooling around it to help the models, but even without that they're definitely capable of designing digital logic.


My guess: it is that those who KNOW the subject realize that LLMs suck at it, and those who do not, do not realize it, since their output is plausible, and sometimes even works.

This actually plays out across every field and is well documented. An expert can recognize the hallucinations and bullshit coming out of LLMs, while non-experts see plausible output and do not know enough to know it is BS.


Wrong. Myself and colleagues know the subject and they are useful in FPGA design. You should stop hallucinating about topics you don't have experience in.


They won't all be fulfilled. I would imagine this is part of the price-discovery process.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: