Vlad Miller

Python private attribute

[Robert Graham @robertgraham]

I read saw Nolan's "Odyssey" on Friday at 3am to 6am.

I couldn't sleep last night, so I read "The Odyssey" for 12 hours.

"What translation?" you ask. "Claude Opus 5" I answer.

I couldn't find a translation in the style I wanted, so I asked the AI to create one to my specifications.

- I don't want an effort trying to fit the english translation into poetic hexameter. I instead want just a literal translation -- just the prose, like in a novel.
- I don't want it updated to modern sensibilities, like Emily Wilson's "translation". That's not a translation but a rewrite. I want to see what the ancient Greeks saw.
- I wanted interruptions and footnotes, where a narrator steps in to explain things that modern readers wouldn't understand, as well as reminders of what they are talking about, such as "Earthbreaker" being synonymous with "Poseidon".

I generated one "book" (chapter) at a time since the AIs can't produce a long document as one step.

The timestamps of the files how how long it took me to read them before I finally dropped asleep in exhaustion.

[Perry E. Metzger @perrymetzger]

No, quantum computers are not about to destroy the security of encryption systems. I have been seeing loads of noise about that, and it’s not true.

To break encryption systems, you need to be able to factor numbers with thousands of digits to break cryptography and we have been stuck with quantum computers being able to factor two digits for 25 years now. 21 is the largest number ever cleanly factored by a quantum computer, and you can do that in your head in a moment, it’s 3 times 7.

This graph is on a log scale (linear in digits), but it would be pretty much the same silly flat line on a linear scale.

You need to be able to handle 1200 digits, but I will settle for being interested again when someone can factor 91 with a quantum computer. Now, a schoolchild can factor 91, it’s 7 x 13, but it would at least represent progress over the current record, which as I said is 21.

(And yes, for the pedants out there, elliptic curve systems are a bit different but it’s the same principle.)

[Vlad Miller @__sattva]

@sult любое видео с канала - за лучшую подачу тем фундаментальной физики, астрофизики и космологии и завораживающий голос нарратора.

https://t.co/RIwGzCqQhb

[James Landrum (no relation) @JamesRLandrum]

@SttyK Fake, there is no escape in North Korea

[Darth Putin @DarthPutinKGB]

Moscow police would’ve prevented this attack from happening if only it was a peaceful protest.

[Perry E. Metzger @perrymetzger]

Today's Starship test was incredibly successful.

Watch as, within a few hours, mainstream media companies print: "Starship disintegrates in atmosphere! Is Elon Musk's SpaceX doomed?"

(They've quite literally done this for every Starship test flight so far.)

[Limaqq @Limaqq_limaqq]

Sorry I've disappeared again, been busy with Ghost in the shell X Deus ex project and finally it's done, check the thread. And thank you so much for your patience #deusex #ghostintheshell

[President Biden Archived @POTUS46Archive]

Aleksey Navalny’s wife, Yulia, and daughter, Dasha, embody his incredible courage.

I was so proud to meet with them and tell them in person: We’re not letting up.

His vision for a brighter future for Russia will live on.

[Safety @Safety]

Our platform's defense mechanism against manipulation and spam mistakenly flagged @yulia_navalnaya as violating our rules. We unsuspended the account as soon as we became aware of the error, and will be updating the defense.

[Дождь @tvrain]

Как российские чиновники комментировали убийства оппозиционеров и чем оправдывают смерть Алексея Навального сейчас — читайте в треде

1/7

[Oliver Carroll @olliecarroll]

Any doubts about whether Navalny was deliberately killed - i.e. acutely forced - two days ago are dissipating. The obfuscation about handing over his body is more than suspicious.

[Perry E. Metzger @perrymetzger]

A couple of times in recent days, I've seen people bring up the notion that consciousness must be some sort of quantum entanglement phenomenon. This seems to be an attempt to smuggle some mechanism into the way the brain functions that humans could never reproduce, in order (it seems) to try to rescue the idea that AIs could never do what brains do.

It's important to understand three things when someone brings this sort of idea up.

The first is that quantum mechanics isn't mysterious or mystical. Understanding it is out of the daily experience of most people, but is not intractable to physicists. Quantum mechanics makes predictions that are no less mathematically based than the predictions of classical physics. If it turned out that the brain was somehow exploiting quantum mechanics, humans could both simulate any such phenomenon on a classical computer (given a speed and power cost) and could build quantum computers to do exactly the same thing (without a speed and power cost).

The second thing to understand is that it is not possible for the behavior of the brain to be an "interesting" quantum entanglement phenomenon. The reasons for this are technical, but suffice it to say, quantum entanglement at significant distances is really, really hard to maintain (thus the difficulty large well funded organizations have had building quantum computers), and a system operating in aqueous solution at body temperature can't maintain entanglement for a long enough period of time to have it exist beyond distances that are microscopic even compared to the sizes of molecules, let alone brain cells. In other words, quantum entanglement can't have anything interesting to do with how the brain works.

It's true that Roger Penrose, who is a really smart mathematical physicist, once speculated that consciousness had such a quantum basis, but it's also true that Isaac Newton, probably one of the greatest physicists of all time, dabbled in alchemy and spent a lot of time on biblical numerology. The fact that you can do the calculation and figure out that entanglement between different parts of the same nerve cell, let alone among different ones, can't happen doesn't seem to have occurred to Penrose but so what, it has occurred to other people, and they figured out it can't happen.

The third thing to understand is, to date, in spite of incredibly intense research on how brains work, neuroscientists have yet to encounter any sort of phenomenon that isn't easily explained by quite ordinary chemistry and physics, without resort to more complicated quantum phenomena. This should not be surprising given that such phenomena seem physically impossible.

Anyway, TL;DR: quantum mechanics is not a good way to introduce mysterious reasons why the function of brains cannot be performed by artificial systems, both because brains don't rely on interesting quantum phenomena and because if they did, we could simulate those too.

[∿ @somewheresy]

AI saved my life.

- I caught a ultra-rare (1.3 in 100k) autoimmune disorder (EGPA) out of nowhere. I used GPT-4 to navigate it in a state full of neanderthals which allow people with rare or hard-to-treat disorders to die at significantly higher rates than the average.

- I found medication, developed with computational biology and machine learning by GSK (mepolizumab) which returned my heart function to normal and controls the propagation of eosinophils in my body by shutting off the IL-5 messenger, preventing them from reproducing. I found the medication by asking a council of LLM API calls possible biotherapies to treat the disorder.

- I prevented myself from bankruptcy by using GPT-4 to negotiate prior authorization requirements, check inattentive doctors' notes, and summarize my lab results for me to learn about tracking and controlling my disease. I further used these tools to negotiate with insurance and credit reporting agencies to prevent myself from getting mired in debt

- I gave myself a new career path by reskilling from an ML hobbyist, to a Prompt Engineer on salary, to an AI Systems Engineer on salary, to Director of AI Systems Engineering on salary, to Head of AI at a different company, allowing me to work 100% remote and asynchronously to have enough income to treat my disease without putting me at risk of dying by being forced into an office full of brain-poisoned boomers.

- I will continue to use these tools in extremely complex ways, including dispatching a farm of agents to augment every single one of my tasks and skills. Now I can run them locally. You cannot stop me from doing this, and I will continue to have them eat whatever material I need to feed them to perform better for their use cases. For "research purposes"

Consider taking control of your life. Consider using these tools to become free

[Sultee @sult]

Нужно отречься от спасительных отговорок: «Как-нибудь обойдется»; «Меня не затронет, я успею отскочить»; «В конце концов, он же не вечен»; «Я уеду туда, где безопасно, куда это не дойдет». Не обойдется, затронет, обязательно дойдет.

https://t.co/pzqXTsMmTk

[Julia Ioffe @juliaioffe]

I always thought Navalny would outlive Putin. I thought good would conquer evil in the end. Either I’m a closet optimist—or Alexey’s optimism was utterly infectious.

[Yaroslav Trofimov @yarotrof]

Let’s see what “devastating” means.

[Медиазона @mediazzzona]

‼️ ФСИН сообщила о смерти Алексея Навального

«16.02.2024 года в исправительной колонии №3 осужденный Навальный А.А. после прогулки почувствовал себя плохо, практически сразу потеряв сознание», — сказано в релизе ведомства

https://t.co/bv2B5bLQzC

[Robert Graham @robertgraham]

People who haven't learned Moore's Law are doomed to re-invent it, like Sam Altman.

Breakthroughs don't happen, as such. Instead, everything is improving by X% per year, like Moore's Law. For chips, X% is high (around 60%). For alternative energy, it's smaller, like 5%.

At such a slow Moore's rate, you don't see progress year to year, but over the long run, it really builds up.

We've already reached the point that bulk energy prices in California go negative during peak sunlight hours of the day because there's overproduction, and they haven't built sufficient storage for it.

What you do is take all the old hardware that is no long economical to run (because modern chips are more power efficient) and build a data center for them in California. Just run them during peak hours of the day when energy is free.

We have multiple exponential growth trends that are compounding every year.
* Moore's Law is making the chips more efficient
* Power is getting steadily cheaper
* AI algorithms are getting steadily more efficient

Altman works at OpenAI where they are able to scale their algorithms inefficiently, just add a $billion more hardware.

Everyone else is constrained on hardware, so are trying to scale algorithms efficiently, to make them better on existing hardware. Nerds like me are tuning 7B models that it on our home graphics card, trying to make them better, rather than using OpenAI APIs that simply make them bigger.

There are no "breakthroughs" happening here. What we have instead is steady growth, but since it's compounding, it'll seem like breakthroughs.

[Linus ✦ Ekenstam @LinusEkenstam]

Milei’s talk at Davos

This never happened.

But with AI (HeyGen) his voice is cloned and turned to English (Elevenlabs) and his mouth is re-synched to match the English words.

This is the true power of AI, anything can be consumed in any language