> Completely bespoke models are typically trained in Python using tools like TensorFlow, JAX or PyTorch that don't have real non-Python alternatives
The article outlines some interesting ways to evade this problem. What's the latest thinking on robustly addressing it, e.g. are there any approaches for executing inference on a tf or pytorch model from within a golang process, no sidecar required?
Practically speaking though, the rate at which models change is so fast that if you opt to go this route, you'll perpetually be lagging behind the state of the art by just a bit. Either you'll be the one implementing the latest improvements or be waiting for the framework to catch up. This is the real value of the sidecar approach: when a new technique comes out (like speculative decoding, for example) you don't need to reimplement it in Go but instead can use the implementation that most other python users will use.
Perhaps check out GoMLX ("an Accelerated ML and Math Framework", there's a lot of scaffolding and it JITs to various backends. Related to that project, I sometimes use GoNB in VSCode, which is Golang notebooks [2].
It is possible to include CPython in a CGO program - allowing Python to be executed from within the CGO process directly. This comes with some complexities - GIL and thread safety in Go routines, complexity of cross-compiling between architectures, overhead in copying values across the FFI, limitations of integrating as a Go module. I am hoping to see a CGO GIL'less Python integration show up here at some point that has all the answers.
These frameworks are C++ under the hood. A far as I know (not too experienced with go) you can use cgo to call any C++ code. So you should be able to serialize the model (torchscript) then run it with libtorch. Tensorflow also similarly has a C++ api
> The valley vortex state presents an emerging domain in condensed matter physics. As a new information carrier with orbital angular momentum, it demonstrates remarkable ability for reliable, non-contact particle manipulation. In this paper, an acoustic system is constructed to exhibit the acoustic valley state, and traps particles using acoustic radiation force generated by the acoustic valley. Considering temperature effect on the acoustic system, the band structures for temperature fluctuations are discussed. An active controlled method for manipulating particle movement is proposed. Numerical simulations confirm that the particles can be captured by acoustic valley state, and can be repositioned through alterations in temperature, while maintaining a constant excitation frequency.
I mean, it _could_ be the job. But a lot of older doctors feel like the younger generation needs to go through the same struggles and hardships they did. This reminds me of something I occasionally see in my parenting of my kids - the desire I have for my kids to do things that validate _my own_ childhood experience.
This line of reasoning is very common. "If young people don't have to endure the same suffering I did then my suffering doesn't mean anything." It comes up in student loan forgiveness, predatory work environments, and many more. Arresting intergenerational harm would admit that ones own suffering was pointless and that isn't something people can deal with accepting. It's a group coping mechanism, which serves a purpose, but also comes with downsides.
I agree with the general principle, but student loan forgiveness is in a completely different category. When a loan is forgiven that's not just free money from the government, it increases the national debt which impacts all citizens. I sympathize with those who made unwise choices as youths to take on excessive student debt. But I sympathize more with working-class people who never took out student loans and instead got a job or enlisted in the military. Why should they now pay for the mistakes of college students who partied for 4+ years in college majors with no earning potential?
Student loan forgiveness also creates a moral hazard. Every future generation of students will then expect that they can take out unlimited loans and if things don't work out then no worries, the debt will just magically vanish.
How do you feel about loan forgiveness for those of us who made good decisions at the time and the conditions changed?
On a personal level, I got MS my very last semester in graduate school. This prevents me from doing the career I trained to do and I instead am working in a lower paying position. Because I can work at all, I'm still on the hook for all my loans even though I can't actually make use of the education I acquired. (Disability discharge is only if you can't work at all, so if you go through all of med school and get a TBI in a car accident the week after you graduate and can't practice but can still work retail, you still have to pay all that money back).
On a societal level, what about the people who were in college during the GFC? I graduated in 2010 and plenty of people graduated in '08/'09/10 who made very reasonable decisions about what to study and how much to take out in loans that got upended because our government and Wall Street can't manage their liabilities. For example, I counted on my father's help to reduce the loans I took out for undergrad, but I ended up with a ton of loans my senior year because everything crashed. Is that my fault or bad decision making? That as a 17 year old I couldn't better predict where the economy would be in 4-5 years than the adults and politicians in charge? Or the kids who've recently gone into CS after it was promised to them as a golden ticket only for the COVID years to fuck them over?
I agree that we shouldn't expect society to be on the hook for bad decisions, but painting student loans as only the result of bad decisions basically looks at the people falling through the cracks and shrugs, declaring them acceptable losses. That doesn't make people want to participate in society or engage in good faith. Which is also a problem.
Yes, I agree. And we shouldn't have paid for the mistakes of GM managers and debtors by bailing them out either. People who make bad decisions should take the consequences.
if you start having kids earlier, you will have more kids.
if so, wont those be more represented in the population and just stablize to closer to where we are today than otherwise?
I wonder if the problem is that, when you give someone advice, you're not giving them the whole thought-structure you've built that strongly implies the advice - you're just giving them the result. So they don't have any of the ideas that support the advice, just a blind aphorism.
Maybe it would help to only give advice in long discussions where you show someone all the structure under the advice and then poke at it together to see where the weak spots are.
This is a theory of mine for which I haven't specifically looked into evidence, so take it with a grain of salt - but I'm convinced that this is why storytelling exists, and specifically why the Hero's journey is so common: stories are essentially beneficial viruses. They exploit our innate desire for discovery by packaging the information into a format that is interesting in the moment, and because we see someone else learning the lesson we are meant to learn, we feel as if we've discovered it ourselves.
This would also neatly explain why we discovered storytelling in the first place, and why so many story elements have changed little over thousands of years - stories "evolved" with us to become more and more effective at inoculating us against dangers which we can survive through our intellect.
> I wonder if the problem is that, when you give someone advice, you're not giving them the whole thought-structure you've built that strongly implies the advice - you're just giving them the result. So they don't have any of the ideas that support the advice, just a blind aphorism.
I'm skeptical. My observation is that it's the messenger that is ignored. Ignoring the advice is simply a side-effect of this.
Consider how many people pay therapists many thousands of dollars to simply hear what their spouse was telling them for years.
Seriously, this was like the biggest epiphany I had out of my 20s. The way to convince anyone of anything isn’t to beat them over the head into submission. It’s to drop the idea off casually, let them ignore it for a while, and watch them slowly come around to this strange idea sitting around that actually seems to work…
The way i've heard this described is "place your truth, gently, next to theirs". Then walk away. No pressure, no sales, nothing. That's the only way it can be adopted.
In that, it's not that the person hasn't heard the arguments before, it's that they haven't met the right person to say it to them. Often, that person is in the mirror.
- If you want money, ask for advice. If you want advice, ask for money.
Do you think people consider recommendation systems (feeds which inform their worldview) to be a part of natural order of things so they don't suspect it. Even if they know it to be algorithmic, nudges can be inserted at appropriate times when our guards are down.
absolutely, but I consider that as the "environment" of cultural evolution so to say
meaning as we collectively get used to a technology we figure these kinds of things out, and even become desensitised to them.
of course people got taken by surprise at first (e.g. first trump campaign's use of those kinds of manipulations over recommendation feeds) but culture (youths and other children) do adapt and react to the "environment" or landscape
This is one of the ways people can go down the rabbit hole of conspiracy theories. They think they "discovered for themselves" some secret truth that "the elite" or "the experts" don't want them to know. Then they find a community of similarly-awakened people who lower their guards even more and point them to other "alternative truths" that they can go DiscoverForThemselves™ too.
Probably the same psychology behind those obnoxious AI-generated ads that say things like "Here's the secret about gut bacteria that doctors don't want you to know!"
This is exactly the reason QAnon took off. It evolved from ARG-ish games without the goal of being persuasive about real-world things. But a participant back then realized how people enjoyed figuring out the 'puzzle', and how to lead them to the conclusion without telling them, and once you add political agendas we got what we have today.
Clicking on the thread I expected this to be the top comment because of how low effort and high brow it is (comments write themselves around pop topics, I find). You can go so far as to skip reading the article. But there is no real way to verify it. It sounds good and gives whoever says it a sense of smugness (“pretty simple, and timeless”), but that’s about it. The person saying it has no idea either way.
> [softbank's offer to rescue wework] is the result of Neumann’s holdup power: Prior to the deal, Neumann is still the company’s controlling shareholder, and he could just say no to a deal that he didn’t like. That might completely evaporate his own wealth, but it would evaporate a whole lot more of SoftBank’s, and it kind of looks like SoftBank blinked first: In effect, the price of Neumann allowing SoftBank to rescue WeWork was that SoftBank had to hand Neumann a billion dollars for himself
The article outlines some interesting ways to evade this problem. What's the latest thinking on robustly addressing it, e.g. are there any approaches for executing inference on a tf or pytorch model from within a golang process, no sidecar required?