It is beyond me how they managed to acquire such funding. Sounds a lot like an earlier quantum computing effort which pivoted to "AI" as well: https://zapataquantum.com
The millstones of the gods grind late, but they grind well. It depends where you're from -- there is always potential for dramatic change in a people as obedient as the German.
I do hope that sufficient humiliation will suffice for DB leadership to enact harsh measures. German precision and punctuality are still there in its people. It's not lost yet.
I've been trying my hand at RL envs for various sparse matrix algorithms in CUDA. It's easy to generate code that "looks good", "novel" and "fast". Escaping the distribution and actually creating novel sequences of instructions or even patterns (has any model come with something as useful as fan-in/fan-out or double buffering patterns that's now ubiquituous?) seems difficult to say the least.
I know some people who took a look at their StaSi files. Today's really different compared to what happened then: There wasn't someone from inside your inner circle transmitting every joke and every utterance to "law enforcement" for them to build kompromat. Sure, nowadays your every move is sold to god-knows-who but this will not impact your ability to choose a course of study (which may or may not suit you academically). You can get a job anywhere provided you have clearance and education needed. You will not look at pictures from your childhood and realize that three out of five people sold you out to get ahead in the queue to acquire a car. Your most intimate thoughts and acts aren't recorded for random bureaucrats to get a laugh out of. Please think again before equivocating "den Realen" with modern democracy. Neoliberal, consumerism societies definitely have their flaws but you're definitely better off on the freedom scale.
Would you happen to have sources on the three orders of magnitude speedup coming at no cost? I'd assume porting + data movement considerations making this task non-trivial.
LFortran is not necessarily using LLVM IR to compile. It's building up an ASR [1] structure that's already being used in the LFortran-specific backends. Potentially it can make full use of Fortran semantics!
Of course it's cancelled, Berset put 1000 people as a hard maximum with events with below 1000 people attending only allowed under risk assessment measures (ie. they might even cancel events below that threshold).
Since it does not treat technologies but rather concepts, it is very much up to date. Everybody in HPC does Monte Carlo and parallel processing, these concepts are still fundamental to working in scientific computing.