I genuinely do not understand why people think my comment on this is a bad take.
I believe it is absolutely ridiculous to have token usage leader boards.
But isn't it also ridiculous for a rational human being to use as many tokens as possible regardless of whether or not that token usage is useful or valuable?
It seems obvious to me that AI service providers make more money the more tokens are used. It also seems obvious to me that companies who sell shovels during a gold rush stand to make a great deal of money, whether or not any gold is found. How is that a good thing writ large?
Anthropic, (and SpaceX), might IPO soon - a move that will grant them access to capital instantly. Capital that comes directly from the retirement plans of all of us.
All I'm saying is if you want to somehow sabotage the game somehow, this is definitely not the right strategy. Political influence would be far more effective.
I'm guessing you aren't invested in any of these stocks. I'm also guessing, to take the perspective of a tokenamxxing GenZ employee, that their bosses are.
Excessive token burning as a tactic to annoy your employer probably does the opposite - it probably makes your employer money.
The more tokens you burn, the more demand for hardware there will be. More demand for hardware means higher stock prices -> more money in your employer's pocket.
The only question is how long that can last. If taken to an extreme, the output of the AI will get worse over time, and if it gets bad enough, for long enough, people will use it less and less, and demand will slowly evaporate.
My point is NOT: I hope this all comes crashing down.
My point IS: tokenmaxing is bad. AND weaponizing it will not have the intended effect, but in fact, it will, in the short term, do the opposite.
Hacker news is not a representative subset of humanity. So you are restricting your range of understanding. Maybe that's necessary for mental health, (it is for me) but the tradeoff isn't great either.
In my mind this is what prototyping is for. Just get it working quickly and see if the concept has legs. But be prepared to completely re-write it because the "just make it work" mindset will make it more difficult to change and improve upon in the future.
But when you factor in today's favorite business model of "make it shitty", perhaps this matters very little.
Yes it is nauseating. And it's the norm. Whenever some company finds a way to make a boat-load of money by exploiting a weakness of human nature, the government will demand a portion of the proceeds. It's business as usual, in the USA at least. Even more nauseating is what they could do with that $2.5B...
Maybe they could use this DDoS attack as their 17th round technical interview. Any candidate who successfully mitigates the attack would then make it to the 18th round. Win win!
I was curious about this so I looked around a bit. My interpretation is that GrapheneOS still has not cracked this nut. Neither has iPhone, unless you enable "Lockdown Mode"
I guess I like the idea of having to think a little more deeply about how memory is used while I'm programming. I know that might sound a bit odd, but before I started learning and using rust, I remember being much less aware of certain things while programming, and then discovering much later that I did something very sub-optimally. So rust is kind of a forcing function in a way - it makes me write better code.
I have also worked on projects (web apps) where GC pauses became annoying to end-users. I wouldn't say this is the fault of GC perse, but writing code without a GC, seems to prevent some performance problems for me, and can also make performance problems stand out sooner in the dev cycle.
Okay I could see that - like implementing linked lists would be easier yeah?
I just feel like not having GC is sort of a deliberate, core design choice for the language. Having strict ownership rules and being forced to think about references feels like a feature not a bug you know? Adding GC feels analogous to the "++" in C++ to me.
Not that I have anything against the efforts people are putting into it though - I'm genuinely curious about what it lets me do better/faster within rust.
I really like rust because it does NOT have garbage collection. Can someone smarter than me help me understand the benefits of having GC in rust specifically? Does it enable things that are more difficult in "non GC" rust?
I would really appreciate it if someone in the know could explain to me how a markov chain with some backpropagation can surpass human cognition. Because right now I call BS.
An LLM is not a suitable substitute for purpose-built SAST software in my opinion. In my experience, they are great at looking at logs, error messages, sifting through test output, and that sort of thing. But I don't think they're going to be too reliable at detecting malware via static analysis. They just aren't built for that.