The very point of the article is that you can win individually and lose as a colective, and that the competitive nature of the field goes against the greater good. And the people betting against AI will be ripped off.
Physics simulation is a dangerous rabbit hole. Even if you focus just on rigid bodies and just physical plausibility there are plenty of open problems related to collision detection and collision resolution. Convex approximations and/or decompositions for geometry and hand tuning of solvers are the norm, balancing robustness and precision against speed.
Reading AI PRs reminds me of Monty Python's holy grenade:
"And the Lord spake, saying, ''First shalt thou take out the Holy Pin. Then shalt thou count to three, no more, no less. Three shall be the number thou shalt count, and the number of the counting shall be three. Four shalt thou not count, neither count thou two, excepting that thou then proceed to three. Five is right out. Once the number three, being the third number, be reached, then lobbest thou thy Holy Hand Grenade of Antioch towards thy foe, who, being naughty in My sight, shall snuff it.'
It reads like those nightmares where you need to pass final exams again.
I guess nowadays it is much simpler to correlate some text with prior work, more so with LLMs. It is like those doping cases where several years later we are able to detect a previously unknown sustance in an old sample.
That's like 4 times the ESA budget, and still insignificant compared to the money poured into AI. Several companies could cover that budget with quarterly profits.
Apple makes excellent hardware (laptop, phone, mini...) to the point I'm willing to pay more for it, but I would prefer a lot to customize my SW. And so I avoid their hardware.
Everybody chooses a favorite depending on their domain.
A function executes, and some error happens:
- Return error value: try to handle the error ASAP. The closer to the error the more detailed the information. Higher probability of recovery. Explicit error code handling throughout the code. Example: maybe you try again in one millisecond because the error is a very low probability but possible event.
- Exception: managing errors requires a high-level overview of the program state. Example: no space left on device, inform the user. You gather the detailed information where the error happened. The information is passed as-is or augmented with more information as it bubbles up the stack until someone decides to take action. Pros: separate error handling and happy path code, cleaner code. Cons: separate error handling and happy path code, unhandled errors.
Worst case scenario: you program in C. You don't have exceptions. You are forbidden to use setjmp because rules. A lot of errors are exposed directly to the programmer because this is a low-level language. You return error codes. Rules force you to handle every possible return code. Your code gets incorporated as an appendix to the Necronomicon.
One reason in Julia for having an organization with multiple repositories is how unnecessary is in Julia to have big packages. It is better to have smaller more focused packages and combine them as necessary. Julia needs to improve some things but I don't think I have found a more modular language.
In some fields throwing away and rewriting is the standard, and it works, more or less. I'm thinking about scientific/engineering software: prototype in Python or Matlab and convert to C or C++ for performance/deployment constraints.
It happens frequently with compilers too.
I think migrating languages is actually more successful than writing second versions.
People adapt to the circumstances. A lot of Python uses are no longer about fast iteration on the REPL. Instead of that we are shipping Python to execute in clusters on very long running jobs or inside servers. It's not only about having to start all over after hours, it's simply that concurrent and distributed execution environments are hostile to interactive programming. Now you can't afford to wait for an exception and launch the debugger in postmortem. Or even if you do it's not very useful.
And now my personal opinion: If we are going the static typing way I would prefer simply to use Scala or similar instead of Python with types. Unfortunately in the same way that high performance languages like C attracts premature optimizers static types attract premature "abstracters" (C++ both). I also think that dynamic languages have the largest libraries for technical merit reasons. Being more "fluid" make them easier to mix. In the long term the ecosystem converges organically on certain interfaces between libraries.
And so here we are with the half baked approach of gradual typing and #type: ignore everywhere.