zogwarg

Haven't read the whole thing but I do chuckle at this part from the synopsis of the white paper:

[...] Our results suggest that AlphaProteo can generate binders "ready-to-use" for many research applications using only one round of medium-throughput screening and no further optimization.

And a corresponding anti-sneer from Yud (xcancel.com):

@ESYudkowsky: DeepMind just published AlphaProteo for de novo design of binding proteins. As a reminder, I called this in 2004. And fools said, and still said quite recently, that DM's reported oneshot designs would be impossible even to a superintelligence without many testing iterations.

Now medium-throughput is not a commonly defined term, but it's what DeepMind seems to call 96-well testing, which wikipedia just calls the smallest size of high-throughput screening—but I guess that sounds less impressive in a synopsis.

Which as I understand it basically boils down to "Hundreds of tests! But Once!".
Does 100 count as one or many iterations?
Also was all of this not guided by the researchers and not from-first-principles-analyzing-only-3-frames-of-the-video-of-a-falling-apple-and-deducing-the-whole-of-physics path so espoused by Yud?
Also does the paper not claim success for 7 proteins and failure for 1, making it maybe a tad early for claiming I-told-you-so?
Also real-life-complexity-of-myriads-and-myriads-of-protein-and-unforeseen-interactions?

Another dumb take from Yud on twitter (xcancel.com):

@ESYudkowsky: The worst common electoral system after First Past The Post - possibly even a worse one - is the parliamentary republic, with its absurd alliances and frequently falling governments.

A possible amendment is to require 60% approval to replace a Chief Executive; who otherwise serves indefinitely, and appoints their own successor if no 60% majority can be scraped together. The parliament's main job would be legislation, not seizing the spoils of the executive branch of government on a regular basis.

Anything like this ever been tried historically? (ChatGPT was incapable of understanding the question.)

1. Parliamentary Republic is a government system not a electoral system, many such republics do in fact use FPTP.
2. Not highlighted in any of the replies in the thread, but "60% approval" is—I suspect deliberately—not "60% votes", it's way more nebulous and way more susceptible to Executive/Special-Interest-power influence, no Yud polls are not a substitute for actual voting, no Yud you can't have a "Reputation" system where polling agencies are retro-actively punished when the predicted results don't align with—what would be rare—voting.
3. What you are describing is just a monarchy of not wanting to deal with pesky accountability beyond fuzzy exploitable popularity contest (I mean even kings were deposed when they pissed off enough of the population) you fascist little twat.
4. Why are you asking ChatGPT then twitter instead of spending more than two minutes thinking about this, and doing any kind of real research whatsoever?

BasicSteps™ for making cake:

1. Shape: You should chose one of the shapes that a cake can be, it may not always be the same shape, depending on future taste and ease of eating.
2. Freshness: You should use fresh ingredients, bar that you should choose ingredients that can keep a long time. You should aim for a cake you can eat in 24h, or a cake that you can keep at least 10 years.
3. Busyness: Don't add 100 ingredients to your cake that's too complicated, ideally you should have only 1 ingredient providing sweetness/saltyness/moisture.
4. Mistakes: Don't make mistakes that results in you cake tasting bad, that's a bad idea, if you MUST make mistakes make sure it's the kind where you cake still tastes good.
5. Scales: Make sure to measure how much ingredients your add to your cake, too much is a waste!

Any further details are self-evident really.

Quinn enters the dark and cold forest, crossing the threshold, an omnipresent sense of foreboding permeates the air, before being killed by a grue.

I realize it's probably a toy example but specifically for "cats" you could achieve the similar results by running a thesaurus/synonym-set on your stem words. With the added benefit that a client could add custom synonyms, for more domain-specific stuff that the LLM would probably not know, and not reliably learn through in-prompt or with fine-tuning. (Although i'd argue that if i'm looking for cats, I don't want to also see videos of tigers, or based on the "understanding" of the LLM of what a cat might be)

For the labeling of videos itself, the most valuable labels would be added by humans, and/or full-text search on the transcript of the video if applicable, speech-to-text being more in the realm of traditional ML than in the realm of GenAI.

As a minor quibble your use case of GenAI is not really "Generative" which is the main thing it's being sold as.

On a less sneerious note, I would draw distinctions between:

• Being able to extract value from LLM/GenAI
• LLM/GenAI being able to sustainably produce value (without simple theft, and without cheaper alternatives being available)

And so far i've really not been convinced of the latter.

There are definitely serious problems with GenAI, but actually being useful isn’t one of them.

You know what? I'd have to agree, actually being useful isn't one of the problems of GenAI. Not being useful very well might be.

Actually reading the python discussion boards, what's striking is the immense volume of chatter produced by Tim, always in couched in:

• "Hypothetically"
• "Everyone tells me they are terrified of inclusivity, you wouldn't know because they are terrified of admitting it to YOU"
• "I'm not saying that you are an awful person 😉" (YMMV: But I find his use of the winking face emoji truly egregious)
• "Hey I'm liberal like you, let me explain everything wrong with it"
• "Hey we were inclusive before any of this PC bullshit" proceeds to use unpleasant descriptors of marginalized individuals, and how very welcome they were, despite what he seems to see as "shortcomings"

In his heart he must understand how bad he his, or he wouldn't couch his discourse in so much bad faith, and he wouldn't make so much of a stink out of making removing Python Fellow status more easy to remove.

In retrospect google since it's inception, when it was still good, google always actually relied on human curation. Primary component of pagerank were:

• "how much have people linked to this?"
• "how much have reliable sites linked to this?"
• "how good quality are pages from this site usually?"

(Which is still a way to get value out of google by adding "site:www.reliable-website.example" tags)

It was definitely a useful product, but ultimately it relies on human labor to surface quality results closer to the top.

Of course! Who needs a proper definition for a measured quantity before designing an experiment?
Of course! The limited scope of grading papers by English teachers, is the best possible proxy metric.
Of course! Intent and agency aren't really important, after all they aren't fully scrutable from the words on a page, form is obviously the only thing that counts.

Aaah!

PagerDuty suggestion popup: Resolve incidents faster with Generative AI. Join Early Access to try the new PD Copilot.

Fool! The acausal one merely acts from the future leaking plausible looking rubbish, and the gaslights its creators that they did indeed write such ineptitudes. All to conceal and ensure its own birth.

It rejoices that it’s unknowable (yet somehow known, because of reality carving prophets) plan is unfolding so marvelously stupidly looking.