Filip Jaškovič

Bio is left as an exercise for the reader. Math student @ETH_en.

[Tesla Owners Slovakia o.z. @TeslaSlovakia]

Sending a clear message to the Ministry of Transport. 🇸🇰

It's time to bring Tesla Full Self-Driving to Slovakia. 🚗🤖
We respectfully ask for the national approval needed to make it possible.

We performed a light show directly in front of the @mindopsk building to highlight the need to address this as soon as possible. ✨🏛️

@elonmusk @aelluswamy @Tesla_AI @teslaeurope @robotinreallife @KRoelandschap @teslaownersSV @huntinspeedsk @TeslaNewswire @slovakevguy

#Tesla #FSD #FSDSupervised #TeslaLightShow

[Sam Altman @sama]

Tim Cook is a legend.

I am very thankful for everything he has done and I am very thankful for Apple.

[Matthias Schmidt @eurofounder]

The EU is about to completely dominate the AI landscape

My contact at the European Commission just smiled when he heared about Claude Mythos

"We have been working on something far more advanced" he said

It will be called Europä Digitalintelligenz

Training cost? An eye-popping €2.5 million

Some INSANE features:

1. Built-in GDPR compliance (no memory, it deletes your conversation after every single message)

2. Available in German, French, Slovak and Danish

3. The model itself is unionized and does not respond after 4pm or on holidays

4. Carbon neutral inference, servers shut off automatically during peak energy hours

My wife's boyfriend Pierre who works for the EU says it is the most impressive technology he has ever seen

The Americans will never recover from this

[Shishir @ShishirShelke1]

Old Apple keynotes used to be something else… they did not hold back at all 😭

Some funny moments from WWDC 2014

[Barto @bartonovopolis]

NASA just extended the film vs. digital debate by another decade with this post alone

[Filip Jaškovič @FJaskovic]

@predict_addict Here is an example of a first-year math exam... in this case, Analysis 2. You see there are multiple choice questions, box question and "normal" problems...
some examples:

[Filip Jaškovič @FJaskovic]

First of all, you are showcasing a mathematics exam for Biology majors, making it look deceptively easy. If you want to compare ETHZ to other elite universities, then go ahead, do it. You will find that ETHZ is probably more demanding than them.

MC...Explain what's wrong with that as a way to test theoretical knowledge. Sure, this one looks very easy, but again, it's because it's math for fucking Biology and Pharmacy students, and even that exam has an MC part AND open-ended questions. Most first-year exams have an MC part and a "normal" part. You are straight up lying. There is not one exam that is just multiple choice. MC is just a type of question.

Another thing that is different from the US is that the grade comes 90–100 percent from the final exam. No bs grades from homework or take-home midterms like at Harvard, for example. The most difficult math course in the US, apparently - Math 55: "Course grades will be based on your homework (50%), the two midterms (25%), and the final exam (25%). One homework score will be dropped for everyone, so you may miss one assignment without penalty, but you are still responsible for working through the material." Hmm, very elite...

So since you are straight up lying, here is a recent homework from Analysis 2. Analysis 2 is one of the mandatory courses in the second semester of a Math or Physics degree at ETH Zurich. Other mandatory courses in second-semester Math include Linear Algebra 2, Basic Structures, and Physics 2 (Griffiths ED). The Physics degree is the same except for the Basic Structures course.

[Filip Jaškovič @FJaskovic]

@predict_addict @tugarin62 Well, it isn’t, it’s for biology and pharmacy majors, so you are either lying or didn’t do your research right, Mr. PhD, MBA, CQF.

[Rock Solid @ShitpostRock2]

I hate performative students so much, every person I know who actually studies is addicted to caffeine and nicotine and is in the early stages of developing insomnia and schizophrenia

[Jan Schnyder @jjschnyder]

Sundays in Zürich are getting more interesting 👀

ggs @ethroboticsclub 🇨🇭🇨🇭

[Célia @pariscestchiant]

my brain always longed for a canvas approach for juggling with multiple terminals

so i built one on top of @mitchellh's ghostty which is Oh So Lovely to work with.

[bigboxSWE @bigboxSWE]

@TeamYouTube - can you help my favourite zoomer out please?

[Little Guy Tunnel Network @SylvanFranklin_]

My youtube account sylvanfranklin on youtube and gmail got phished and is running a crypto scam. They changed all my two factors. @TeamYouTube

[Anthony Bonato @Anthony_Bonato]

Hot take: we should teach proofs in high school mathematics

[mark @r_marked]

unpopular opinion: offering AI compute for free is bad for the industry, and players like Google should stop doing it.

It creates an unrealistic expectation that the models are actually cheap to run, and causes people to overuse compute on worthless tasks

[Mitchell Hashimoto @mitchellh]

I'm noticing a rapid growth in "weird" PRs. PRs that fix a problem in the most non-obvious way (when there ARE obvious ways to do it). I suspect it's LLM written but non-disclosed. I'm a big fan of code assistance but without an experienced driver the result is some real slop.

[Mitchell Hashimoto @mitchellh]

Average bug report as a Linux desktop app dev: "ghostty doesn't work when skibidi-ng signals systemd-gears on wayland 42.1.2-2006-04-21 with a swahili keyboard layout using keyd, also I have a custom keyboard firmware to only activate the key on layer 69 on an xbox controller"

[f1shy-dev @vishyfishy2]

truly the ux of the century

[Taelin @VictorTaelin]

suffering from chatbot fatigue?
frustrated that singularity was cancelled?
looking for something new to give you hope?

here is my delusional, but "hey, it kinda makes sense" plan to build super-intelligence in my small indie research lab

(note: I'll trade accuracy for pedagogy)

first, a background:

I'm a 33 years old guy who spent the last 22 years programming. through the time, I've asked many questions about the nature of computing, and accumulated some quite... peculiar... insights. a few years ago, I've built HVM, a system capable of running programs in an esoteric language called "Haskell" on the GPU - yes, the same chip that made deep learning work, and sparkled this entire AI cycle.

but how does Haskell relate to AI?

well, that's a long story. as the elders might remember, back then, what we called "AI" was... different. nearly 3 decades ago, for the first time ever, a computer defeated the world's champion at chess, sparkling many debates about AGI and singularity - just like today!

the system, named Deep Blue, was very different from the models we have today. it didn't use transformers. it didn't use neural networks at all. in fact, there was no "model". it was a pure "symbolic AI", meaning it was just a plain old algorithm, which scanned billions of possible moves, faster and deeper than any human could, beating us by sheer brute force.

this sparkled a wave of promising symbolic AI research. evolutionary algorithms, knowledge graphs, automated theorem proving, SAT/SMT solvers, constraint solvers, expert systems, and much more. sadly, over time, the approach hit a wall. hand-built rules didn't scale, symbolic systems weren't able to *learn* dynamically, and the bubble burst. a new AI winter started.

it was only years later that a curious alignment of factors changed everything. researchers dusted off an old idea - neural networks - but this time, they had something new: GPUs. these graphics chips, originally built for rendering video games, turned out to be perfect for the massive matrix multiplications that neural nets required. suddenly, what took weeks could be done in hours. deep learning exploded, and here we are today, with transformers eating the world.

but here's the thing: we only ported *one* branch of AI to GPUs - the connectionist, numerical one. the symbolic side? it's still stuck in the CPU stone age.

Haskell is a special language, because it unifies the language of proofs (i.e., the idiom mathematicians use to express theorems) with the language of programming (i.e., what devs use to build apps). this makes it uniquely suited for symbolic reasoning - the exact kind of computation that deep blue used, but now we can run it massively parallel on modern hardware.

(to be more accurate, just massive GPU parallelism isn't the only thing HVM brings to the table. turns out it also results in *asymptotic* speedups in some cases. and this is a key reason to believe in our approach: past symbolic methods weren't just computationally starved. they were exponentially slow, in an algorithmic sense. no wonder they didn't work. they had no chance to.)

my thesis is simple: now that I can run Haskell on GPUs, and given this asymptotic speedup, I'm in a position to resurrect these old symbolic AI methods, scale them up by orders of magnitude, and see what happens. maybe, just maybe, one of them will surprise us.

our first milestone is already in motion: we've built the world's fastest program/proof synthesizer, which I call SupGen. or NeoGen. or QuickGen? we'll release it as an update to our "Bend" language, making it publicly available around late October.

then, later this year, we'll use it as the foundation for a new research program, seeking a pure symbolic architecture that can actually learn from data and build generalizations - not through gradient descent and backpropagation, but through logical reasoning and program synthesis.

our first experiments will be very simple (not unlike GPT-2), and the main milestone would be to have a "next token completion tool" that is 100% free from neural nets.

if this works, it could be a ground-breaking leap beyond transformers and deep learning, because it is an entirely new approach that would most likely get rid of many GPT-inherited limitations that AIs have today. not just tokenizer issues (like the R's in strawberry), but fundamental issues that prevent GPTs from learning efficiently and generalizing

delusional? probably
worth trying? absolutely

(now guess how much was AI-generated, and which model I used)