Elsewhere: Twitter Bot Wars & GPT-4's True Power

Posts I did in other venues on topics of interest to my newsletter audience.

I hate to be like “web3 fixes this” about anything in 2023, not because it’s not true but because nobody wants to read that now that they’ve either lost their shirt in crypto or are feeling smug about others having lost their shirts in crypto.

But still, when Musk started tweeting about how the future of social media is paid subscriptions because botnet operators will be priced out, I had to say something. Crypto has been working on this proof-of-human problem for a long time now, and there are shipping solutions.

You can read my argument in

RETURN

, or in Meteor, a new web3-focused publication where my RETURN piece was reprinted.

Musk Is Right About AI Bot Swarms, But We Can Fix Social Without Users Paying

The other piece I did this week outside this newsletter was a post for

Pirate Wires

where I look into the issue of GPT-4’s missing parameter count and what it might tell us about the true power that OpenAI is holding back:

With GPT-4, OpenAI Is Deliberately Slow Walking To AGI

This piece is a follow-up to my most recent post on the nascent AI culture wars, and it picks up where that piece’s postscript left off. A sample:

Here’s my thesis: The performance numbers published in the GPT-4 technical report aren’t really like normal benchmarks of a new, leading-edge technical product, where a company builds the highest-performing version it can and then releases benchmarks as an indicator of success and market dominance. Rather, these numbers were selected in advance by the OpenAI team as numbers the public could handle, and that wouldn’t be too disruptive for society. They said, in essence, “for GPT-4, we will release a model with these specific scores and no higher. That way, everyone can get used to this level of performance before we dial it up another notch with the next version.”

… So based on this paragraph alone, we should look at GPT-4’s benchmark performance as a pre-selected outcome. They looked at a point on their parameters vs. performance curves and said, “Let’s turn the scaling dial so that GPT-4 lands… there! That’s about what we estimate society will be ready for when we launch this in a few months.”

If you’re a regular reader of this newsletter then you’re going to want to read the whole thing because it’s kind of important and relevant to the whole issue of how fast all of this is going.

Finally, I made substantial editorial contributions to this large

RETURN

explainer on central bank digital currencies (CBDCs), which I encourage you to check out if you're interested in crypto, money, and related issues:

CBDCs In Their Own Words: Technocrats Blackpill Us On The Future Of Money

Loose threads

I’m not sure how many of you follow me on the bird site, so here’s a poll asking whether periodic roundups of some of my more substantial threads are worth doing in this newsletter.

So with this first thread, I kinda lost my mind a bit… I signed up for Midjourney and just couldn’t stop with this silliness.

jonstokes.(eth|com) @jonst0kes

I finally ponied up for Midjourney access. It's going about like you'd expect.

jonstokes.(eth|com) @jonst0kes

The thing that's not widely appreciated enough about programmer-oriented advances like this, is that they increase the second derivate -- the rate of change of the rate of change. All the new ML-powered coding tools are 2nd derivative maximizers. 🚀🚀🚀

Lior⚡ @AlphaSignalAI

JUST IN: @stripe partners up with @OpenAI to enhance their documentation: GPT-powered Stripe Docs. I believe this will quickly become a standard. https://t.co/4mCfoWPExu

jonstokes.(eth|com) @jonst0kes

He's describing how OpenAI'S ChatGPT plugins integrate with your service. You just describe what your API does in English, and OpenAI's model figures out how to call it and make it work with ChatGPT.

Mitchell Hashimoto @mitchellh

For those who aren't aware: you write an OpenAPI manifest for your API, use human language descriptions for everything, and that's it. You let the model figure out how to auth, chain calls, process data in between, format it for viewing, etc. There's absolutely zero glue code.

jonstokes.(eth|com) @jonst0kes

Me: Well grandson, when I was your age, humans wrote the software. GS: Oh wow? How was that even possible? Me: It kinda wasn’t possible. Software was just terrible. Nearly all of it was garbage. If AI hadn’t arrived in time, software would’ve killed us. GS: We survived an… https://t.co/82tNI4WVWT

jonstokes.(eth|com) @jonst0kes

The lightning storm of awesome (jonstokes.com/p/i-say-this-u…) strikes again. This but for all forms of electronically mediated symbol manipulation labor, and in a very compressed timeframe. And then in a slightly more extended timeframe, for all forms of labor.

Autism Capital 🧩 @AutismCapital

This will continue to be a very common story moving forward. The fastest to accept, process grief, and evolve, will thrive. https://t.co/VwvKAARiyN

jonstokes.(eth|com) @jonst0kes

People are pattern-matching the current AI hype to past hype cycles. This is an error. I've lived & worked through all the hype cycles back to the dotcom boom/bust, & AI has a quality that sets it apart from all the others: the sense of crisis.

jonstokes.(eth|com) @jonst0kes

Was recently talking to an HFT guy who does cycle-by-cycle optimization in C++, & I realized there are whole pools of untapped code optimization talent in our civilization that can & will be shifted to the emerging market for inference & training optimization. Gonna be wild.

Justine Tunney @JustineTunney

Yesterday my changes to the LLaMA C++ file format were approved. Here's how folks in the community have been reacting to my work. https://t.co/cLzg19n81f https://t.co/5gcrVp8jvn