Alexandros Marinos 🏴‍☠️ Profile picture
Founder of Aspiring practical philosopher. Systems thinking is the way to a positive-sum civilization. #gameb #OCCUPYUNIVERSE
Hecate's Crossroad #QVArmy Profile picture Karen Hamilton ❌ Profile picture Secret Squirrel 🇺🇲 Profile picture phal Profile picture Still Learning Profile picture 68 added to My Authors
Mar 26 15 tweets 4 min read
OK, let's do a thought experiment together. Help me think this through.

If generative AIs like ChatGPT/GPT-4 and the like could be seen as a kind of requirements-to-source compiler, what would software projects look like? I'm going to ask you to just accept as an assumption that these kinds of models can be turned into reliable-enough compilers, but first, I'll give you some reasons for my optimism:

1. Reflexion for GPT-4 already does pretty well at generating code:
Mar 26 6 tweets 3 min read
Enough moaning, let's do some work.

Let's figure out how many lines of rapid improvement in favor of GPT and related LLMs are open right now.

I'll list as many as I can think of, please reply with as many more as you have, and I'll append the ones that make sense here. The obvious one is more parameters. And by more, I mean by orders of magnitude. Take what worked for the GPT-2/3/4 transitions and do more of it. A lot more. Image
Mar 22 25 tweets 5 min read
An update on my ChatGPT/LLM thoughts. Not claiming any particular expertise, but maybe these thoughts will trigger responses that help better understand what's coming.

First and foremost I feel like understanding this thing is the most important thing I can be doing. As far as I can tell, this technology can provide 95% of a correct solution on any problem humans can already solve. And it can provide that solution on-demand, and fast.

For domains where it can be nudged to correct its errors, that 95% becomes effectively 100%.
Mar 21 14 tweets 5 min read
Soon: Twitter Space w/ @mormo_music to discuss AI risk, safety, & what part of rationalist discourse can be rescued.

All welcome, speakers will be mostly rationalist-adjacents who were on covid heterodox side. (Yes, labels suck).

Preliminary thoughts:🧵… AI capabilities are exploding almost as fast as AI discourse. In all this, I notice a wave of alarmism from my former tribe, the rationalists.

I am sympathetic in principle, but during the pandemic I saw behaviors and reasoning that makes me not take their conclusions seriously
Mar 21 4 tweets 1 min read
If the rationalists were remotely serious about preventing AI risk, they would have defended humanity's sensemaking as sacrosanct, as we will need high quality communication and collective reasoning in a time of emergency.

They didn't, because they arent. I suppose this is worth considering. I would say "not sufficiently serious", since clearly other concerns have entered the picture, but I do realize the "no true scottsman" vibes.
Mar 18 9 tweets 3 min read
ChatGPT is FIRE. Image @drrollergator is this right or am I missing something? I was sort of expecting the inverse result, which is bad enough at 30% false positives, but maybe I'm missing something?
Mar 17 26 tweets 10 min read

Does Sam Harris have an expert "smell test" or did COVID cause him loss of smell (and taste)?

By picking experts to defer to, he can't be blamed for their errors.

But only if his smell test can tell deep state stink apart from devoted scientist scent.

How did he do? Well, Harris has had Renee DiResta on, twice, approvingly. (Episode #145 & #310). Most recently to argue against the value of the Twitter files. Today she was exposed as being former CIA, and leading the Virality Project, which drove COVID censorship...
Mar 17 8 tweets 2 min read
If you value my recommendations at all, stop what you're doing and watch this ASAP.

No matter how incompetent the "experts" seem, things can always be much worse. The reviews are in.
Mar 16 19 tweets 5 min read
Elon should continue to fire leakers and haters until "platformer" stops writing bad faith commentary on sources it doesn't share with us. Let's walk through this piece and see what is known about the incident: So, first of all, even the article itself, (in the 13th paragraph..) admits that there was a legitimate technical issue that the engineers were trying to resolve.

And yet in the next paragraph it puts the words "the problem" in quotes. Bad faith overflows.
Mar 16 7 tweets 3 min read
The problem with large scale interventions like masking is that to the degree they work, they also present an evolutionary challenge for the virus to mutate around, since they favor more infectious variants.

Did the complexity theorists have that in their models? In other words, under the assumption masks worked, we had a measure we could have deployed carefully to protect frontline workers and the elderly, and we used it up to protect healthy toddlers who stood to gain nothing from it.
Mar 14 39 tweets 12 min read
Starting a thread on this because .. (insert reason here).

Let's see where it goes, and how long I make it through.

Lex Fridman, Sam Harris, 4 hours, here we go.

THREAD Hey, 14 minutes in before he starts talking about Trump! Good Job Sam! And surprisingly little to comment on.
Mar 14 4 tweets 1 min read
I suspect bifidobacteria are about to become a recurring topic.

Example 1: Example 2:…
Mar 13 14 tweets 4 min read
Here are a bunch of bad SVB takes to avoid:
1. This is a bailout of big banks
2. This is a bailout of big tech
3. This is a bailout of bad management
4. This is a bailout at all
5. This is a taxpayer expense in favor of the few
6. This is another abuse of the money printer.

What…… It's not a bailout of big banks because if they didn't step in, all the money would have moved to the big 4 "systemic" banks, blowing up the rest.
Mar 10 7 tweets 3 min read
The greatest sequence of videos in the history of YouTube... for a certain value of "greatest".

Step 1. Guy reverse engineers how to make paint that actually cools the surface it's on below ambient temperature:
Step 2. Other guy finds how to do it even better, solving problem guy 1 had:
Mar 8 7 tweets 2 min read
We just survived(?) a pandemic of conspiracies.

Send me your favorite, 100% documented conspiracies we've learned about during the pandemic.

Here's a few of mine: Double whammy: Employees conspiring to blackmail their boss during a pandemic, and Fauci conspiring to cut off the head of the Centres for DISEASE CONTROL AND PREVENTION from participating in meetings about a PANDEMIC. NBD.
Mar 8 8 tweets 4 min read
🧵I recently started digging into the proposed deal between @dailywireplus and @scrowder.

While people focused on the drama, nobody was fully analyzing the terms.

So I got the text and did the kind of analysis I would do if the deal was offered to me.🧵…… @dailywireplus @scrowder A few points that people seem to have missed.

1. It really is a contract. It contains a binding section that makes the rest of the term sheet also somewhat contractual.
Mar 7 14 tweets 6 min read
OK Twitter, I need some help.

Can we list all the ivermectin RCTs that the @GidMK / @K_Sheldrick / etc team has confirmed they have had access to raw data (aka. IPD) for?

I wrote my piece below based on my recollection, but forgot I have my collective intelligence access back! First, Biber, as confirmed by Kyle Sheldrick on PubPeer:…
Mar 3 24 tweets 8 min read
Twitter has a 🇩🇪 problem (aka. what might have happened to my twitter account, and what could happen to yours) 🧵

Let's talk about how the NetzDG law, and others like it, puts platforms in an impossible position.

p.s. I'm livestreaming the…… But first, what (might have) happened to my Twitter account. On Jan 11, I received an email from Twitter telling me that my (perfectly innocuous) tweet was reported but (thank Thor!) I was not found guilty.

(I discovered this email just yesterday.) Image
Mar 3 9 tweets 2 min read
Magnets! How do they work?

(Also, I'm back! Miss me?) It's not magic, it's made of magnets, steel, plastic, (and tape)

Basically a crappy version of this:…
Jan 16 4 tweets 3 min read
I just started looking into the case of Maddie De Garay. It seems there is complete blackout on her story.

Given that her case brings into question Pfizer's child vaccine trial and therefore its approval, which allows Pfizer to secure immunity, I'd expect it would be addressed? The amount of propaganda in this piece is off the charts. All meta no meat. I feel dirty just reading it. How could a human being put these words to paper?
Jan 16 7 tweets 2 min read
Indeed. The public health establishment lying through its teeth and putting a gun to people's heads has created so much bad blood that we risk a future *actually effective* intervention being negatively received. Public health has one resource only: trust. And they squandered it. Here's a sample of all the times I've been making this exact point over the years, in case people feel like claiming I'm saying this just now: