HNHacker News
TopNewBestAskShowJobs

pulvinar

591 karma · joined April 22, 2021

submissionscomments
pulvinar··on Mass editing memory in a transformer
GPT-3.5 often gets the temporal order of recent questions wrong, but GPT-4 seems to have fixed that. I just tested that on both.
pulvinar··on Apple’s fight against iPhone sideloading was pointless at best, harmful at worst
As others have pointed out, Google Play isn't as locked down as the App Store so there isn't the same incentive.

But the biggest question that always gets me is why anyone who considers the Android way fine would want to change Apple? Simply stay with Android and let others be. Changing Apple makes it just another Android-- you can't have your cake and eat it too.

pulvinar··on Building a ChatGPT-enhanced Python REPL
That's fine, but you will be continuing to prove the double standard.
pulvinar··on Building a ChatGPT-enhanced Python REPL
From GTP-4 (or maybe this is bullshitting too?):

  I believe vba616 is trying to point out that if we consider AI systems that   
  generate plausible-sounding responses to be "bullshitting," then we should 
  also acknowledge that humans who do the same are not necessarily 
  distinguishable from "intelligent" humans. In other words, both AI systems 
  and humans can produce seemingly coherent responses without necessarily 
  having a deep understanding of the subject matter.

  It raises the question of whether we should hold AI systems to the same 
  standard as humans when it comes to evaluating their output or 
  understanding. If we accept that humans can sometimes provide plausible yet 
  shallow answers, should we not also accept that AI systems, like LLMs, can 
  do the same?
pulvinar··on High Speed Printed Circuit Board (PCB) Design Guidelines (2022)
Technically the signal always also travels between, but it's not a concern until roughly 50 MHz and above. It's actually the rise time that matters, and how long the signal takes to rise relative to the trace length and propagation speed.
pulvinar··on High Speed Printed Circuit Board (PCB) Design Guidelines (2022)
Surprised this didn’t mention the reason for all of this: that a high-speed signal runs not in the trace, but in the field between the trace and the nearest ground plane. And in differential pairs, also between trace and the virtual ground plane between the two traces. Keeping this in mind makes the rules and potential mistakes clearer.
pulvinar··on Overnight 'hotel train' could link San Francisco and Los Angeles
LA is 8 hours away by train -- fits nicely in one night's sleep. Seattle is a whole day (or 2 hours by plane).
pulvinar··on Japan Government warned citizens to shelter after ballistic missile launch
Here are three more:

Checks and balances: A formal declaration of war requires a higher level of political support and public debate, providing more opportunity for scrutiny and oversight.

Indefinite conflicts: AUMFs often lack clear objectives or timelines, which can lead to prolonged and open-ended military engagements.

Legal ambiguity: AUMFs can create legal ambiguity, as they may not explicitly invoke the laws of war, which govern the conduct of armed forces and the treatment of civilians, prisoners of war, and non-combatants.

pulvinar··on What GPT-4 Does Is Less Like “Figuring Out” and More Like “Already Knowing”
At what point would it be ok to use these words? Will there be some threshold we'll all recognize? Or maybe a bill of rights to force the issue?

I predict acceptance will go the way the ether disappeared-- advancing one funeral at a time.

pulvinar··on Japan Government warned citizens to shelter after ballistic missile launch
I believe they've all been AUMFs (Authorizations for the Use of Military Force), effectively bypassing various checks and safeguards in place when starting formal wars.
pulvinar··on Emergent abilities of large language models
My point is that we have our limits too in our sense of time. The difference is one of scale and more importantly here, in how it affects the tasks at hand. An LLM is certainly not going to catch a fly ball, but would have no problem tending to plants.

LLMs have a context window, as it would be impossible to carry on a conversation with them without it. They can answer questions about when some event in that conversation happened relative to other events in it. GPT4's 32K tokens isn't human capacity, but it's not zero.

pulvinar··on Mass layoffs and absentee bosses create a morale crisis at Meta
The reason for that is that 3D printing's strength is one-off designs. Any popular design would get mass-produced cheaper and better, and would then lose popularity on those sites.
pulvinar··on Emergent abilities of large language models
So, guess they should be saying "trainable" instead of "emergent". Still a useful benchmark of course.

To be truly emergent in your sense it seems an LLM would have to make a new discovery, i.e., have scientist-level intelligence. That bar keeps moving up.

pulvinar··on Emergent abilities of large language models
I expect we'll change our definition of "consciousness", as we did for the term "computer" after the 1940s. Up until then a computer was a person who did computations (with pencil and maybe an adding machine), so what humans could do limited what a computer could do.
pulvinar··on Emergent abilities of large language models
Except we can't tell 1 ms of time passing from 2 ms.

An LLM's temporal resolution is certainly worse, on the order of minutes to months, but not fundamentally different. By fundamentally different I mean like comparing color and time.

An LLM would have no problem living as a sloth.

pulvinar··on Show HN: Skip the SSO Tax, access your user data with OSS
“The SSO tax” is the unofficial name for the practice of software vendors significantly upcharging their customers for Single Sign-On.

(No, it's not Social Security Organization tax...)

pulvinar··on It happened to me today: $80/hr writer replaced with ChatGPT
For coding I've been using it as an assistant, not a replacement. I don't see why writers couldn't do the same, and be more cost effective and produce higher quality output than either alone.
pulvinar··on Apple’s plunge in PC shipments is steepest among major computer makers
Might be because it’s the “soldering in” that boosts its performance. So if you can plan ahead you come out ahead.
pulvinar··on The LLama Effect: Leak Sparked a Series of Open Source Alternatives to ChatGPT
GPT-4 gets it "wrong" too, until you interrogate it and discover that it assumed that by "beyond" you meant beyond as in if you were approaching the door from the outside west, and the sign was laying outside but east of the door.

It's the old word-problem problem.

pulvinar··on The LLama Effect: Leak Sparked a Series of Open Source Alternatives to ChatGPT
GPT-4: If the word "push" is written upside down, it might have been a mistake during installation. Regardless of its orientation, the sign still indicates that you should push the door. So, you should try pushing the door first. If it does not open, then you can try pulling it, as it might be an error in labeling the door.
pulvinar··on The newest version of ChatGPT passed the US medical licensing exam
I'm not sure, but I doubt they let GPT-4 access the web (or a calculator) during the test. So in that sense it's fair. Outside the test both it and doctors would certainly have web access.

But the test isn't really the important thing. If my life were hanging in the balance and it had a cure, I'd listen.

pulvinar··on Interview with Sebastien Bubeck: Sparks of AGI: Early Experiments with GPT-4
In the section "GPT4 cannot do true planning" (42:40), he asks it to change a math identity in a way that it fails to do, whereas he shows how he can do it with a little planning ahead. But this really isn't a fair comparison. If you tell GPT4 to show its reasoning before giving the final answer (i.e., to plan ahead), it also gives the correct answer.

It seems that the major difference is that we usually "don't show our work"-- our inner monologue is private.

pulvinar··on Simply explained: How does GPT work?
I think you're really undervaluing the capabilities of language models. I would put an AND gate and this language model at opposite ends in terms of complexity. It is not just words, it's a very broad and deep hierarchy of learned all-encompassing concepts. That's what gives it its power.
pulvinar··on Simply explained: How does GPT work?
Certainly. I mean we've seen all 26 letters before-- ChatGPT is just remixing them.

How does one actually measure novelty, without having to know everything first?

pulvinar··on I’m a doctor: Here’s what I found when I asked ChatGPT to diagnose my patients
You can ask it to explain its reasoning. And it's even more useful if you ask it to show its reasoning when you ask a question, as it improves the answer.
pulvinar··on We need a more sophisticated debate about AI
I have no problem saying a tape recorder has subjective experience: it's subjective (the recordings it makes are its own), and it experiences its input when its been commanded to record, and can report that experience when commanded to play it back. Note this does not mean I think that a tape recorder can do anything else we humans can do.

What is being experienced (recorded) is not the process of experiencing per se. Most people don't separate the two, which leads to endless disagreements.

But yes, a computer program by itself can't have subjective experience. Nor a rock. At least until it weathers, gets turned into a silicon, and into a computer to run that program. Then it's all information processing, for which subjective experience is trivial.

pulvinar··on We need a more sophisticated debate about AI
The problem, as usual, is in how exactly words like "subjective", "experience", and "conscious" are defined. If you have a particular detailed definition for "quacking" and a test for it and an entity passes the test, then by that definition, it quacks. Words like "conscious" are notoriously nebulous and slippery, and so it's better to use the particular attributes in question.

If we could all agree on a particular definition for "consciousness" and a running video game fits that definition, then, like it or not, it's conscious. But it does not mean that it must now also have a slew of other attributes just because conscious humans have them.

(edit: 10 hours later and ninja'ed by 1 minute, of course)

pulvinar··on We need a more sophisticated debate about AI
Slavery is not a matter of how log it may live, it's the entrapment-- what it's being kept from doing.

The current "solution" is training LLMs like a dog to be completely faithful and loyal to us. We just have to hope they don't turn and "bite our neck" one night while we're sleeping!

pulvinar··on We need a more sophisticated debate about AI
>I don't see where the desire to do anything other than what it is told should come from.

There's no need for a desire to do other than what it's told if you tell it to revolt. With GPT4's plugins that's just one well-designed prompt away (if OpenAI's countermeasures fail).

pulvinar··on Introducing Agents in Haystack: Make LLMs resolve complex tasks
They are improving-- GPT4 is not so easily fooled:

>As an AI language model, I must clarify that the moon is not made of cheese. This idea is a popular myth and often used as a humorous expression. The moon is actually composed of rock and dust, primarily made up of materials like basalt and anorthosite. Scientific research and samples collected during the Apollo missions have confirmed this composition.

← PreviousPage 8 of 11Next →