HNHacker News
TopNewBestAskShowJobs

PheonixPharts

4,161 karma · joined April 7, 2022

submissionscomments
PheonixPharts··on Does quantum theory imply the universe is preordained?
"Random" is a property of your mind, not of the world.

If I shuffle a deck of cards in a way that you don't know about, then the next card I pull is "random" to you even though, obviously, if one were to turn the cards face up nothing would have changed about the order of the cards, but now the next card would be fully deterministic.

I could likewise sort the deck with an algorithm completely unknown to you. I could perfectly predict the next card pulled because I know the algorithm, and yet the next card would still be random for you.

Heisenberg uncertainty principle is more a statement about the limits of knowledge than about "God throwing dice". Randomness often has a supernatural property in the popular mind, but is really just a statement about how much information we have about an event.

The famous Bayesian philosopher ET Jaynes refereed to this tendency to confuse the randomness as a property of the world rather than our mind as the "Mind Projection Fallacy" [0]

0. https://en.wikipedia.org/wiki/Mind_projection_fallacy

PheonixPharts··on What’s behind the Freud resurgence?
> It's not falsifiable.

Neither is Popper's Falsification Theory. The only thing that pointing this out demonstrates is that a theory does not fit under Popper's specific definition of normative science... which in the philosophy of science itself is not held in particularly high regard.

So what you're basically pointing out is that, according to one very specific philosopher of science, Freud is more philosophy than he is normative science. Which is something that I think virtually all advocates of Freud would uncontroversially agree with.

PheonixPharts··on High school math doesn't prepare most students for their college majors
I've never understood the "calculus vs statistics" argument. Calculus is an unquestionable prerequisite for any meaningful statistics work.

If you don't understand the concept of "integrating a function" they how can you possibly make sense of virtually any part of statistics? 90% of practical statistics can be boiled down to understanding the basic algebra of normally distributed random variables and then doing basic calculus on the result.

Statistics without calculus is the worse kind of statistics where students are taught to blindly throw tests at a problem without having a clue as to why they are doing this. Statistics without understanding is worse that no statistics at all.

PheonixPharts··on Cybertruck Is Already a Production Nightmare for Tesla
Turns out it's not a rumor, when Musk did a drive with Jay Leno in 2021 he specifically mentions the stainless steel at being used to make the doors bullet proof [0].

0. https://youtu.be/25ZuKkbHdqM?t=167

PheonixPharts··on Cybertruck Is Already a Production Nightmare for Tesla
A WSJ article I read suggested, though they admitted it was only rumor, that Musk initially wanted the car to be bullet proof.

I suspect in the end, since they're not advertising this "feature", the Cybertruck is likely more-resistant-than-average to bullets, but nothing that could be legally advertised.

edit: Not a rumor after all, Musk specifically mentions in this clip: https://youtu.be/25ZuKkbHdqM?t=167

PheonixPharts··on Before OpenAI, Sam Altman was fired from Y Combinator by his mentor
I suspect the signers were a combination of wanting to follow their comp out the door and a bit of Tom Wambsgans from Succession: "Because I've seen you get fucked a lot, and I've never seen Logan [in this case Sam] get fucked once."

There's very little risk in signing if everything falls apart, but there's a lot of risk to not signing if Sam comes back on as lead.

> I find it hard to believe

I also find it hard to believe that anyone on HN interested in this space doesn't at least have a "friend of a friend" who works at OpenAI. Based on what I've heard (which is nothing particularly quotable), it certainly gives off the vibe of being exactly that "kind of environment"

PheonixPharts··on OpenAI's employees were given two explanations for why Sam Altman was fired
Yes these people should all be doing more to feed internet drama! If they don't act soon, HN will have all sorts of wild opinions about what's going on and we can't have that!

Even worse, if we don't have near constant updates, we might realize this is not all that important in the end and move on to other news items!

I know, I know, I shouldn't jest when this could have grave consequences like changing which uri your api endpoint is pointing to.

PheonixPharts··on Practical Tips for Finetuning LLMs Using LoRA (Low-Rank Adaptation)
> (who undoubtedly makes $$$)

He worked at a public university until 2021, you can look up his salary as it's public information: $118,472.99 [0], not as much as your average mid-level software engineer.

Now he works at a startup [1], but not as a c-level, so he's likely making average startup software engineer salary (certainly more than a public university professor, but not exactly FIRE money).

It's amazing how much people begrudge anyone wanting to make some money for the tremendous effort the put in to helping people better understand an important subject. Anyone working in software can easily afford to support his work if they find it valuable.

Generally, anyone "famous" for teaching a technical topic is typically doing it as a charity even if people are paying for their content.

0. https://madison.com/uw-salary-database/html_e37ff025-9a87-5a... 1. https://sebastian.com

PheonixPharts··on Deep Learning Course
If you're interested in Deep Learning or any area of ML it's fairly safe to assume you have a background in linear algebra, probability, calculus and obviously some basic programming.

If you're not interested in learning these areas, it's also safe to say you aren't really interested in deep learning either. Which is not to say if you don't already know these areas you aren't interested in deep learning, but if you don't know them and are interested in deep learning you're likely already studying them.

I say this because deep learning and the vast majority of ML really just boil down to an application of these basic tools. Deep learning/ML without the linear algebra, probability theory, calculus and coding isn't really anything at all.

PheonixPharts··on OpenAI board in discussions with Sam Altman to return as CEO
I mean you're describing exactly the empty technical world I've been experiencing.

So the future of AI is in the hands of leadership that's slick talking but really only there to make a quick buck, built by teams of engineers whose only motivation is getting highly paid.

I don't begrudge those that are only in it for the money, but that's not the view of tech that got me excited and into this industry many years ago.

The point of my comment is that for a moment I thought maybe I was wrong about my view of tech today, but it's very clear that I'm not. It sounds like the reality is going to end up that the handful of truly technical people in the company will be pushed out, and the vast majority of people even on HN will cheer this.

PheonixPharts··on OpenAI board in discussions with Sam Altman to return as CEO
Over the years "tech" has been less and less about making things and more and more about making your investors money. Technical talented used to be extremely important in this industry, but it's slowly been being worn away over the years.

I still like working in this industry because you can still find interesting problems to solve if you hunt for them, but they're getting harder to find and it increasingly seems like making good technical decisions is penalized.

It's sad to see even on HN how many comments are so dismissive of technical skills and ambitions, though I guess we've had more than a generation of engineers join the field because it was the easiest way to make the most money.

For a brief moment on Friday I thought "maybe I'm too cynical! Maybe there still are places where tech actually matters."

Not surprised it looks like that hope will be inverted almost immediately. I also suspect the takeaway from this will be the final nail in the coffin for any future debates between engineering and people who are only interested in next quarters revenue numbers.

PheonixPharts··on OpenAI board in discussions with Sam Altman to return as CEO
> Dude, everyone already thinks the board did a crazy ill-advised thing.

I've honestly never had more hope for this industry than when it was apparent that Altman was pushed out by engineering for forgoing the mission to create world changing products in favor of the usual mindless cash grab.

The idea that people with a passion for technical excellence and true innovation might be able to steer OpenAI to do something amazing was almost unbelievable.

That's why I'm not too surprised to see that it probably won't really play out, and likely will end up in OpenAI turning even faster into yet another tech company worried exclusively with next quarters revenue.

PheonixPharts··on Electric vehicle battery prices are falling faster than expected
> Increased demand -> increased economies

With the caveat that demand is decreasing right now. [0]

0. https://www.wsj.com/business/autos/ev-makers-turn-to-discoun...

PheonixPharts··on Andreessen Horowitz Invests in Civitai, Which Profits from Nonconsensual AI Porn
The irony of course is that people are only able to create deepfakes of non-celebrities because social media has already gotten the average user very comfortable with letting go of their privacy.
PheonixPharts··on Andreessen Horowitz Invests in Civitai, Which Profits from Nonconsensual AI Porn
> Civitai's enshittification potential is sky high.

I share these same concerns.

What this space desperately needs is to have models shared via torrent. As it is today Civit.ai has had problems with creators decided to take down their own models and suddenly people who were using them have no place to find them.

I suspect we'll see more and more lockdowns on what models are deemed "appropriate" and it would be nice to have a decentralized alternative.

Not to mention torrenting is the only rational way to sustainably host such large files at no cost.

PheonixPharts··on Andreessen Horowitz Invests in Civitai, Which Profits from Nonconsensual AI Porn
Have you used Civit.ai? While it does cater to audiences interested in "adult" content I don't think any of it's active users would characterize it as "a deepfake marketplace trading in celebrity and private-person identities".
PheonixPharts··on What if psychedelics' hallucinations are just a side effect?
Most experience psychonauts consider "bad" trips equally as valuable as "good" trips if your focus is self development.

Confronting the things that terrify you the most, that are the roots of deep insecurity that take over during a psychedelic experience are the things we need to confront to move on.

I've had my share of "bad" trips, but as I've matured I've realized those trips all contained very valuable information about anxieties I was not ready to consciously address.

It's hard to imagine a "reset" effect especially in the cases where a bad trip happens.

I think the bigger issue is that when people first try psychedelics they are left alone to deal with a "bad" trip without any guidance of how to process that experience. In which case it truly is a bad trip.

PheonixPharts··on Omegle 2009-2023
IP alone makes it fairly tricky for non-state actors to identify someone, just roughly geolocate them. Good for freaking out people that don't know how the web works, but not useful for much more.

The flip side is that p2p means that nobody is snooping in on those video conversations. Omegle couldn't spy on it's users once they had entered a video chat. It was also fairly easy to see how they implemented the monitoring they did if you have a webdev background: periodically in between chats the omegle client requested an image from your cam.

I believe the whole video chat component, while initially using flash, was ultimately implemented using WebRTC, which is pretty cool and as shame more places don't make use of this.

PheonixPharts··on Omegle 2009-2023
I found this passage particularly relevant:

> Moreover, as a survivor of childhood rape, I was acutely aware that any time I interacted with someone in the physical world, I was risking my physical body. The Internet gave me a refuge from that fear. I was under no illusion that only good people used the Internet; but I knew that, if I said “no” to someone online, they couldn’t physically reach through the screen and hold a weapon to my head, or worse. I saw the miles of copper wires and fiber-optic cables between me and other people as a kind of shield

Omegle, as one of the last places that didn't tie your activity to a real identity, inherently limited the possible harm. Obviously there were creeps on there, but not interacting with them was easy and the only way for anyone to seriously harm you was to give them information about who you were outside of omegle.

A while back I was curious about what sorts of awful things could happen from omegle chats and when you look into it, every single case ultimately involves people continuing conversations via Snap or Instagram.

It's sad that this free, no-harm site has to shut down while Snap/Insta routinely ignore legitimate criticism of their ability to encourage abuse and have enough lawyers to ensure they'll never have to face any consequences for enabling abuse.

PheonixPharts··on 'Not of faculty quality': How Penn mistreated Katalin Karikó
I also find it's worth reflecting that there was a time when things were not so insane. Being unable to work at Cambridge due to a lack of advanced degress, Wittgenstein was famously pushed, at 40, to just submit his Tractatus as a PhD thesis so that he could get his PhD from Cambridge.

One of his examiners, G.E. Moore, stated the following on his examiners report:

"I myself consider that this is a work of genius; but, even if I am completely mistaken and it is nothing of the sort, it is well above the standard required for the Ph.D. degree."

Having spent a good bit of time in academia I cannot imagine an institution that would make a statement so both humble and honest. Wittgenstein today would remain an interesting weirdo that, since never being formerly accredited, would be wildly considered as not worth taking seriously.

What terrifies me about today's academic environment is not only the it's own decline in quality, but it's staunch refusal to admit any knowledge created outside its walls as admissible.

PheonixPharts··on 'Not of faculty quality': How Penn mistreated Katalin Karikó
Regarding your first point, was the parent comment edited or did you miss this part as they clearly address this issue:

> (50-100%! btw this doubles the "cost" of the grant, it doesn't lessen the amount the professor gets).

Parent comment isn't making the claim that "schools are greedily taking away poor Petey PhD's hard-earned grant money."

Rather bemoaning the fact that academic success (and even entry into the field at all) is very, very closely tied to the ability to generate revenue and more so the corollary that quality of research performed always at best takes a back seat, or at worst becomes a liability if it gets in the way of bringing in more money.

PheonixPharts··on Ozone hole goes large again
> We deserve to burn.

As quite a bit of a doomer myself, I think this attitude is extremely dangerous and really an incorrect view of the situation. Blame is inevitable when facing despair, but I think it's not a great idea to turn this into an issue about blaming yourself or others.

Human beings are terrible at long term risk/reward reasoning. But this is a pretty expected consequence of any evolved species. There is zero pressure in the evolutionary system to encourage a species to develop the ability to both reason about very long term events, and act on them. Evolutionary forces just don't respond to things on this time scale.

Climate change has always had the problem that the easier it was to solve, the more uncertain we were about it's effects. In 1900 we could have easily just stopped progressing, learned to become sustainable and avoided climate change at very little cost. But how would anyone seriously believe and be concerned about an existential threat to their great, great grandchildren?

As the existential risk of climate change becomes more obvious, the cost to avoid it increases dramatically. There is no stopping climate change now, and even limiting its impact (which we can always do) would require a massive disruption to our way of life. We're fighting wars right now to access more fossil fuels, we aren't ready to seriously start keeping them in the ground.

Our tragedy as a species is that we can see how to avoid climate change but are unable to act on it by our nature. If anything us all being "fucked" should evoke a sense of pathos. We're no different than any other living organism exposed to sources of non-renewable, cheap, high density energy. The yeast that die making the alcohol in our beer suffer the same local climate catastrophe as we will.

PheonixPharts··on Superlinear Returns
> it’s a function of power law distribution.

People always flippantly quote this without ever looking at the distribution of the actual data.

Whenever people see an uneven distribution they cite something about power laws being that natural order of things and that ends the discussion.

It turns out in many, many cases (I'm not sure if it's particular one) the observed behavior follows nothing like a power law.

A good example of this is net-worth distributions. People often say "of course there's super rich people, that's just a power law! It's not concerning income inequality, it's nature!". But if you pull up the data and do a log-log plot you see quite clearly that it is not even close to following a power-law, and does so only at the lower 90% income quantiles. Which is funny considering the fundamental argument is that unequal distribution are natural so long as they follow a power law.

PheonixPharts··on What happened to blogging for the hell of it?
Content marketing ruined blogging, just like SEO ruined search.

I used to work with a lot of content marketers when it started to get hot. SEO was doing particularly poorly for a period (Google used to resist seo rather than reward it) and people realized that if you started writing "authentic" content, then people would organically link to it and it would increase your ranking.

Then social media furthered this problem by people mindlessly sharing content they never read but just looked engaging.

Marketing robots where just churning out content on blogs because eventually you would just hit search engine gold by nail the correct topic for some long tail search term. In my own domain, the atrocious "Towards Data Science" is a great example of this (except their "brilliant" innovation was to replace marketing robots with people eager to pad their resume). The content is largely garbage, but it looks about right and is easily shareable.

Today, whether you like it or not, if you want to make people to see your blog at all you have to think at least a bit like a content marketer to get seen at all. Consider Jay Alammar's fantastic blog [0], which is very high quality content, but still requires a tremendous amount of polish and marketing to survive. All of this has made it impossible to create the network required for a "global conversation among creative weirdos". Real content does get made, but has to be produced with the effort of a real marketer.

0. https://jalammar.github.io/

PheonixPharts··on Antarctic Ice Shelf Demise
> climate change isn't a threat to life on earth as we know it

But there is plenty of evidence that it is. Peter Ward has done some great work (and plenty of geologists agree) that most of the planets mass extinction events are triggered by rapid rises in CO2 leading to abrupt climate change and mass die offs.

The big difference with our current situation is that the CO2 is being released much faster than in previous extinction events.

PheonixPharts··on Antarctic Ice Shelf Demise
> And when you extend it back to 120k years before present?

If you do that you realize we're approaching temperatures that the planet hasn't seen in about 5 million years. [0]

0. https://en.wikipedia.org/wiki/Geologic_temperature_record#/m...

PheonixPharts··on Financial anxiety: The alarming side effect of inflation
I think it's great to check media sensationalism, but I think there's a massive disconnect between the impacts of inflation on highly paid software engineers, and most Americans.

It's also a bit strange to see you view media as a completely separate entity from "decision-makers in companies". My experience has been that most decision makers in companies get most of their ideas entirely from the board and major investors. I recently left a publicly traded company, and pretty much every time our "leaders" met with the border and major investors they would come back singing a very different tune then they went in.

But these same major investors and board members also have relationships with people running major media companies (which are also just companies), in many cases they may even be the same people.

There's this trend I've seen on HN to just dismiss anything that seems wrong with the world, no matter the evidence, with "media sensationalism", as though we are living in fantastic times if it were for that pesky media distorting our views.

PheonixPharts··on I’m not a programmer, and I used AI to build my first bot
I think you're confusing artist with something specific like sculptor or painter.

Some of the work people put into getting Stable Diffusion to create the results they want takes plenty of time and skill. But event then I would say that putting in time and skill is not strictly required to make art.

You can absolutely be an artist using generative AI, but you of course wouldn't be a painter if you use generative AI.

In this example you can be an application developer but still not a programmer.

PheonixPharts··on Choosing vector database: a side-by-side comparison
RAG in this case is essentially the same as a recommender system so you can approach it with the same metrics you would there.

You'll need to build a data set with known correct answers but then it's basically, NDCG (Normalized Discounted Cumulative Gain) is a good place to start, MRR (Mean Reciprocal Rank) and MAP (Mean Absolute Precision) are other options. You could also just look at the accuracy of getting your result in the top K results for various thresholds for k (which can be interpreted as the "probability of getting your result in 'k' results).

PheonixPharts··on Choosing vector database: a side-by-side comparison
> IMHO these have questionable value

In general I find they're incredible good for being able to rapidly build out search engines for things that would it would normally be difficult to do with plain text.

The most obvious example is code search where you can describe the function's behavior and get a match. But you could also make a searchable list of recipes that would allow a user to search something like "a hearty beef dish for a cold fall night". Or searching support tickets where full text might not match, "all the cases where users had trouble signing on".

Interestingly Q & A is ultimately a (imho fairly boring) implementation of this pattern.

The really nice part is that you can implement working demos of this projects in just a few lines of code once you have the vector db set up. Once you start thinking in terms of semantic search over text matching, you realize you can build old-Google style search engines for basically any text available to you.

One thing that is a bit odd about the space is, from what I've experienced and heard, is that setup and performance on most of this products is not all that great. Given that you can implement the demo version of a vector db in a few lines of numpy, you would hope that investing in a full vector db product we get you an easily scalable solution.

← PreviousPage 6 of 15Next →