Ioannis Mavroukakis

I think car door handles should still work when the car is on fire

20 χρόνια Elf and Safety δεν φεύγουν έτσι απλά με αυτή την χώρα 🤣

Το να έχεις μια μπαταρία λιθίου πάνω από heat source? Α για τον ύπνο μιλάμε !

Arrakis, one of the most complicated knitting patterns available on Ravelry, is finished. Hellshawl is done.

When Wanting Never Stops: A Humanist View of Schopenhauer and Billionaires - TheHumanist.com "Success does not end wanting, and therefore cannot be trusted to regulate itself."

thehumanist.com/commentary/w...

In our current democracy, the loss of institutional memory, the distain of expertise, and the celebration of ignorance frighten me.

Democracy dies in darkness
– Washington Post

Also

Will the last person leaving please turn out the lights?
– Washington Post

🥹

Maybe my lazy ass can sit down and learn jj this time 🥹

He’s a goddamn treasure.

YouTube video by The Pragmatic Engineer The third golden age of software engineering – thanks to AI, with Grady Booch

As expected, @booch.com is an immensely enjoyable guest www.youtube.com/watch?v=OfMA...

🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣🤣👏

The supporting, testable, repeatable evidence simply does not exist.

Are LLM architectures useful?

Certainly.

Are they sufficient for building a general intelligence?

I see considerable evidence to the contrary.

Are they part of such a solution?

Very likely.

But “a team of top researchers” has said so in Nature is a vapid appeal to authority. There are a multitude of dissenting “top researchers” around the world.

Extraordinary claims demand extraordinary evidence, as Carl Sagan observes.

I don’t have a one-track mind. I have an eight-track mind.

It gets tangled easily, fades out in the middle of a thought, and is quickly wearing out.

The Pitt got greenlit for a season 3, out next February, and the announcement has very funny comment from HBO chairman Casey Bloys that’s like “yeah, turns out if you just make a TV show like how we used to make TV shows you can have a 15+ episode season every year 🤷‍♂️”

Would still eat. I will be so upset when I can’t get all the citrus.

I made a few of mistakes, typical 2nd attempt arrogance- at this temp, it needs the full 45 minutes for my oven.
I also realised that the recipe calls for lining the _bottom_ and I lined the whole pan 🥴
I also forgot to spray oil 🥹

vibecoded web apps have such boring security bugs. "the whole database was wide open". oh. ok.

at least have some class and write some sql-injectable php. maybe a little stack buffer overflow as a treat.

Dialled the oven down 15C as its fan assisted,

A winter citrus upside down cake, slices of citrus arranged in a random fashion.

Send help.

I’d say the stakes are a little higher now, than a layout decision.

Ah I see we have arrived at the “it’s your fault stage”.

i get that the stereotype is a hacker in the basement using kali in their free time to be a big serious pentester but uh. can we normalize people having 1) hobbies outside of work, & 2) using a variety of tools instead of the flashiest ones that tech influencers pretend they use

That’s quite the “if” and I fear what will take for that gatekeeping to be torn down, is another dot com bust (old enough to have lived through the first one). Thus far, LLM have managed to flood the creative arts, as Deezer discovered and drive the price of RAM through the roof. Wonder what’s next.

I hope you are right. I don’t believe the machine code vs 3GL as an abstraction is directly applicable to an LLM’s output, especially without several safeguards and possibly, formal methods of verification.

My concern is, that we are riding a huge hype cycle that will do more harm than the (eventual?) good it purports it will. I certainly have very little faith in the tech bros that front the major players.

I agree, but given that trajectory, are we producing “engineers” that will be unable to judge the output? Sure, you might be able to define the problem but can you really trust what the LLM produces if you cannot decipher it? Is the hype for AI on the same scale as the move to 3GL?

How AI assistance impacts the formation of coding skills Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

Perhaps. Perhaps not. www.anthropic.com/research/AI-...