CS unrolls an ancient scroll

This is great news of applied computer science and potential great impact [nod to Matias πŸ˜‰] on future archeological work in deciphering the history.

https://scrollprize.org/firstletters

For those who haven’t been following, an ancient scroll has been meticulously scanned using various methods, and the data has been released as open source.

A cumulative prize of $1 million is up for grabs for anyone who can successfully interpret the text from the available data. And the best part? Everything is being done in a truly open-source spirit.

And even greater is that anyone can still be involved or just learn about, as everything is open source.

So for anyone wanting to go into applied machine learning, algorithms and data processing, this is a great way in πŸ™‚

If you’re keen on updates and insights, I recommend following Nat Friedman on x-Twitter. He offers quality insights on the progress and more. πŸ™‚

Safety critical Rust (finally) qualified

Rust is one of the prime candidates for safety-critical systems.

And it looks like it is finally coming of age πŸ™‚

After long and arduous work, Ferrous systems have released open source safety-critical qualified Rust compiler.
Check below announcement for more information:

https://ferrous-systems.com/blog/ferrocene-open-source/

But what is qualified tool

Those who are involved in safety critical work β€” and I claim no expertise here β€” know that one of the bigger ‘pains’ is to have qualified safety-critical tools to use.

For a ‘regular’ developer most of the tools are taken for granted like compilers, operating systems, frameworks, etc. Bugs and undefined behaviors1 are expected to exist and even used as ‘optimizations’2.

While determinism, quality, and reliability are desired, they often take a back seat to features, speed of delivery, and user experience.

But if you work on safety critical systems, it is vice-versa.
One of your biggest needs is a proof of deterministic behavior.

And there is a simple analogy.
If a houseβ€” or even more critically, a skyscraper β€” lacks a stable foundation, it’s only a matter of time before something catastrophic occurs.
You need and want proof (as much as feasible) that your foundations are stable. And you generally want that reputable company with reputable build process makes your foundations.

That is what qualification means in practice.
You want a tool that you can rely on where you will not be nasty surprised in a worst possible moment3.

And now Rust got its ‘reputable’ badge for use in safety critical systems.

Speed-up move to modern languages

There is also another reason why I’m Optimistic on Rust and new modern languages.

C and C++ are great languages for their uses, performance and low-level. C is a glorified assembly, and C++ gives you million ways to write bad or wrong code and only few good ones.

To use C and C++ in safety critical environment, you are severely restricted how you can apply those languages. Just check MISRA or AUTOSAR4 rules for writing automotive-grade quality code.

And even then you must use different set of tools (static and dynamic analyzers, linters, etc) just to make sure that you haven’t done offset +1 in memory access, or to implicitly do wrong typecast.

Long time ago I considered myself as a ‘quite good’ at C++, but then the more I learned, the less I knew 5 πŸ˜€

If you need to put so many restrictions and constantly invent different ways how to stop people making almost invisible mistakes in a language, that means that you need to reconsider do you really need to step into that mine-field6.

And all of the hoops significantly slow the actual development time.

So lets see what future will bring7 πŸ™‚

  1. https://en.cppreference.com/w/c/language/behavior β†©οΈŽ
  2. https://alexpolt.github.io/undefined.html β†©οΈŽ
  3. https://www.motortrend.com/news/nhtsa-tesla-autopilot-investigation-shutoff-crash/ β†©οΈŽ
  4. https://www.autosar.org/news-events/detail?tx_news_pi1%5Baction%5D=detail&tx_news_pi1%5Bcontroller%5D=News&tx_news_pi1%5Bnews%5D=39&cHash=e4f521f7b674bdfd7c1fade308cf2ea8 β†©οΈŽ
  5. “Within C++, there is a much smaller and cleaner language struggling to get out” – Bjarne Stroustrup β†©οΈŽ
  6. Of course, C and C++ are highly used languages and have a great applications (just look in CERN or Linux kernel). But they should be not and are not applicable everywhere as the learning curve is high and space for errors even higher β†©οΈŽ
  7. Greetings to team Trust at VCC doing great work for production grade Rust in actual car πŸ™‚ β†©οΈŽ

Beauty of Zenbleed and its walk-through

Last week (23w30) I stumbled upon such a ‘great’ vulnerability, but also the great explanation (and I’ve read a ‘few’).

Explanation of Zenbleed is here https://lock.cmpxchg8b.com/zenbleed.html.
And I highly recommend it to anyone interested in learning some of the ‘magic’ of modern CPUs on a ever-expanding 40+ year old x86 instruction set.

And a bit of jealous how clean and clear the walk-through is written πŸ™‚

Old is new again (at least on the web)

I came across a comprehensive analysis of all the β€˜inovations’ in full-stack development, which can be found at https://www.bitecode.dev/p/hype-cycles. It brought back some fun memories πŸ™‚

Every few years, I found myself needing or wanting to create a website. This involved trying to use “best” tools and frameworks for the job that everyone was hyping at the time

Yet, each time, I found myself gravitating back to good old WordPress, with a sprinkle of jQuery or a lightweight framework.

Why?

Well, I felt β€˜stupid’ that it took so much time to set up a simple CRUD website using these supposedly “modern” tools.

The rule: simple task needs simple tools, always πŸ™‚

George Carlin VS state-of-the-art AI

My current take on recent AI development is that it is getting more and more useful.

BUT, in the end, it is still a glorified, but novel statistics [1].

And whenever AI learns from human input, I find funny/insightful/dark quote from George Carlin:

Or in other words, current crop of AI cannot escape law of large numbers [3].

Especially in today’s world, you need very-very-very valid input sources vetted, by of course, error prone humans.
Even reviews for the simplest of stuff you can buy cannot be trusted as they are bought in bulk.
And as misinformation efforts are running loose in the wild, it is hard to keep sources clean.

Even for pure technical domains such as programming, ChatGPT has been banned by StackOverflow [2] due to high percentage of only looking correct answers.

And some fun chat-bot responses from history characters πŸ™‚

https://twitter.com/RealESonneborn/status/1615794316504440839

Conclusion

So until AI can learn different real-life models instead of shoveling data into hundreds of billions of (statistical, black-box) parameters with insane compute power needs, it will just be useful statistics.

But there is a lot of room for research in how even today’s AI works, even on much smaller scale models.

References

[1] – https://towardsdatascience.com/no-machine-learning-is-not-just-glorified-statistics-26d3952234e3

[2] – https://meta.stackoverflow.com/questions/421831/temporary-policy-chatgpt-is-banned

[3] – https://www.investopedia.com/terms/l/lawoflargenumbers.asp