Content About Stocktwits
No Image Available

'What’s in a Name?' From SpaceXAI to 'J-Space' to China AIs in US. ARD #113”

Author
Michael Parekh AI: Reset to Zero
12 hours ago
14 MIN READ

Today’s theme: what’s in a name? The names and definitions we’re all constantly reaching for across AI. Three events for the AI Tech Wave — SpaceXAI’s rebrand, Anthropic’s language for how Claude’s ‘mind’ works, and the US-China open-vs-closed model split — each with my Take first, then my Overall Take.

(1) SpaceXAI — The Rebrand From xAI

MP TAKE: It’s been a long road built by Elon with ever-changing stories: from Twitter, to X, to xAI, to SpaceX/xAI, to now simply SpaceXAI — new logo and all. All the while still in orbit around Tesla — named, of course, for the man who bested Edison on electricity over a century ago — and all of it still driven by Elon Musk on his cadence.

Every one of these name changes points at the same trajectory: Terafabs and Space AI data centers, with a lot of other AI objectives here on Earth along the way — Optimus humanoid robots and Robotaxis, in the tens of millions and beyond. It’s all driven by Elon’s vision of how AI can propel him toward his North Star, and by how his never-ending narratives keep driving his boundless ambitions forward — evolving, always, to a spot marked ‘X.’ Which was the key name all along, with him leading the way. A digital Pied Piper of the 26th century.

Sources, in narrative order: Business Insider — xAI makes its rebrand to SpaceXAI complete with a new logo. For longtime readers, in narrative order: ‘SpaceX/xAI IPO filing outlines Elon’s boundless AI Ambitions’ in AI-RTZ #1066; and ‘Elon Musk boosts SpaceX/xAI pre-IPO story with “Terafab”’ in AI-RTZ #1034.

(2) Anthropic’s Work on Claude’s ‘J-Space’

MP TAKE: This is one of the best mainstream explanations I’ve seen of how AI models actually work. Anthropic has taken the language humans use to convert ideas into ‘reasoning,’ and shown its bare approximation in the matrix math running on Nvidia’s AI GPUs — and it does the job in words and in a genuinely good 5½-minute YouTube video (with excellent images). It really cuts to why this probabilistic AI is so different from the deterministic computing we’ve had for over 60 years. When I explain AI to mainstream folks, this is the part everyone nods at but nobody quite gets, because there’s math involved. ‘J-Space’ moves the ball forward.

And here’s the deeper point it illustrates: language is only an approximate gateway. How our brains really work is beyond language — as many scientists have theorized, and we still barely understand our own brains. We’ve somehow converted that into massive amounts of matrix math, calculated on hardware and software that costs hundreds of billions of dollars — and that’s what makes this stuff work. Once you truly grasp it, you see the delineation between machines and humans: we should not be anthropomorphizing the machines, even though they seem to be thinking, seem to be reasoning. Those are human words — names — roughly applied to the matrix math. Sometimes it leads us astray, into thinking these things are sentient. They’re not. They’re doing incredible, still-surprising things — but far beyond ‘stochastic parrots’ too. The map is not the territory.

Sources, in narrative order: Anthropic — A global workspace in language models · “What’s at the center of Claude’s mind?” (YouTube explainer, ~5½ min — a must-watch). Also, Axios on “Claude has carved out itw own space to ponder”. For longtime readers: ‘Anthropic Claude app takes the lead over OpenAI ChatGPT’ in AI-RTZ #1014.

(3) No Language Barriers Between US and Chinese AI Models

MP TAKE: Two things can be true at the same time, as I repeat often on AI-RTZ. On one side, OpenAI and Anthropic will keep charging premium à la carte prices for their top frontier models — once the government lets them run free and sell aggressively worldwide — even as their aggregate prices fall through software and compute efficiencies. And businesses and developers will keep paying, in the billions. Anthropic is already at a $50 billion revenue run-rate as of last month; between the two, they’re plausibly on a $100 billion-plus run-rate within 12 months, in time for their trillion-dollar-plus mega-AI IPOs. They’ll capture the majority of the economics of the AI-model business for at least the next two or three years.

Conversely — and at the same time — open-source models (from China especially, but also from US players like Nvidia and Apple) will take the majority of the usage. Once enterprises figure out which applications actually work, they move those to the open-source models: cheaper, non-metered, runnable on local and data-center compute — with advantages beyond price, in customization and in building privacy and IP protections around their core assets. Remember: open source drives 80-90% of everything we’ve done in computing for decades — the internet itself is mostly open source. So separate the two: the percentage of value captured on an economic basis versus the percentage of actual compute usage. The frontier labs take the economics; the open-source models take the compute. Opposite ideas and definitions, happening at once.

Sources, in narrative order: CNBC — Chinese AI models are gaining ground with US companies as OpenAI, Anthropic costs surge. For longtime readers, in narrative order: ‘US Companies building on Chinese open-source LLM AI Models’ in AI-RTZ #923; ‘It’s Business and Personal in OpenAI/Anthropic pricing battles’ in AI-RTZ #1048; and ‘How Nvidia & Apple can be the global US Open Source AI Champions vs China’ in AI-RTZ #1089.

MP OVERALL TAKE

Today’s theme — ‘what’s in a name?’, eloquently described in its wide and deep meanings by Shakespeare— runs the whole gamut of AI definitions of names and ideas. It’s the branding at Elon’s SpaceX, a single name now stretched across a complex, fast-evolving portfolio of AI products and services — digital, physical, and eventually out there in Space.

It’s the deep innards of how AI models actually work — dubbed ‘J-Space’: language fused with deep matrix-math concepts, still built on how we think human brains work in a neural context, itself only barely understood. The attempts — by Anthropic and all the AI companies — to name and explain those innards matter, because that’s how we get to interpret, explain, audit, and monitor these systems: how we keep them within guardrails, safe and secure, and protect the world from cybersecurity risks. Not just for AI researchers, but for everyone.

And finally, it’s how all of that translates into the financial language of costs and prices, as the gap between the best frontier models and the more prosaic open-source ones keeps widening — even as both can be popular at once, and both grow aggressively over the next two or three years, until this all matures and gets figured out. These are the useful core names for the realities driving the AI Tech Wave. Get the names right, and you can think more clearly about what’s actually happening underneath them.

Gadget AI — The Privacy We Haven’t Yet Defined for AI Smart Glasses

MP Take: Over the next couple of years we’ll see millions of AI smart glasses — not just from Meta, but Google, Amazon, and Apple too — plus a whole family of AI wearables, especially the ones with cameras. The Verge had another great piece on this. And there’s a long history here, going back two decades, all inspired by nearly a century of science fiction. Remember Google Glass a decade ago — it looked amazing, had real promise, and then collapsed on societal concerns about creepiness, cameras, and spying. These have mostly been technologies for technology’s sake, still looking for their true product-market fit, with a huge amount of social and user-experience work yet to be done — even as they sell in the millions, at hundreds or thousands of dollars each.

That sounds like a lot, until you remember we live in a world with over five billion smartphone users out of eight billion people — so the numbers get very big, and so does the experimentation. We’ve barely begun defining the privacy here — across dozens of companies, an escalating profusion of regulations (federal, state and local, country by country), and the bottom-up question of where our privacy lines lie, where they stretch, and where they truly stop. This time the cameras are always on — so the definitions have to come first. But I’m optimistic it’s a net positive over time, as the technologies get better at helping us execute our sci-fi-driven visions.

Sources, in narrative order: The Verge — “I Spy” (smart glasses, AI wearables, surveillance and privacy) · Everything spy movies get right (and wrong) about smart glasses. Gizmodo — How Science Fiction Ruined Smart Glasses (video essay — a must-watch). For longtime readers: ‘Meta Leans in on AI Smart Glasses Science Projects’ in AI-RTZ #849.

Questions

Q1 — What is MP most looking forward to in camera-driven AI capabilities, on glasses or AirPods?

Contextualized daily journaling of everyday activity — especially on a private basis — translated into usable data on our daily lives. This is the new data: the stuff we live with every day that never gets captured in any database of record, or in the language models we use, closed or open. That’s the valuable thing worth experimenting for — to somehow capture it, and then leverage it in the next iterations of AI, whether driven by the big models in data centers or smaller models running locally on our devices. While respecting the privacy and trust boundaries of others — somehow.

Q2 — What does Michael worry about with that same capability?

That we end up with billions of AI Agents running around carrying the personal context of millions and billions of people — mostly misunderstood and misused. How do regulators deal with it? How much of our privacy — our expectations once that data is captured in the digital domain, across the world — gets eroded? These are huge issues that won’t be solved overnight with gadget A versus gadget B; they get figured out by society over time, on 10- and 20-year curves. The real question: what is truly gained for the privacy lost? As said above, I’m still directionally optimistic on a net positive result over time.

Wrapping up

Today’s AI-RTZ #1140 — A post-mortem on Anthropic’s ‘Blip 2.0’ with the US Government — a look back at the roughly three-week government stoppage of Anthropic’s latest Fable and Mythos models, now slowly being lifted. In my view ‘Blip 2.0’ is still continuing — there are more speed bumps and speed-limit changes ahead for the frontier models this year, and they’re worth watching closely for a whole set of reasons I’ve been tracking. Recommended as today’s reading.

Tomorrow — ARD 114 on AI-RTZ 1141.

Thanks for joining us today, AI Curious Folk. Stay tuned.

Full Source Reading —

For the broader context, see the canonical sources for ARD 113 — in today’s narrative order:

Theme — “What’s in a Name?”

Event 1 — SpaceXAI Rebrand From xAI

Event 2 — Anthropic’s Claude ‘J-Space’

Event 3 — US and Chinese AI Models

Gadget AI — AI Smart Glasses Privacy

Clips from today

Clip 1 — AI Wearables: Privacy Concerns & Market Potential

Watch on YouTube Shorts

Millions of AI smart glasses and camera-equipped wearables are coming from Meta, Google, Amazon and Apple — but the privacy questions are barely defined.

MP Take: After two decades of building smart glasses ‘just because we could’ — and watching Google Glass collapse on creepiness concerns — this time the cameras are always on. The definitions have to come first. Still, I’m optimistic it’s a net positive over time.

Clip 2 — AI’s New Data Frontier

Watch on YouTube Shorts

The most valuable data isn’t in the language models — it’s the everyday context of our lives that never gets captured in any database of record.

MP Take: What I most want from camera-driven AI is contextualized, private journaling of everyday activity — turned into usable data for future remembering and doing more, leveraged by the next iterations of AI, in the cloud or locally. That’s the new data frontier.

Clip 3 — AI: Beyond Language, Beyond Sentience

Watch on YouTube Shorts

Anthropic’s ‘J-Space’ is one of the best mainstream explanations of how AI models really work — probabilistic matrix math, not the deterministic computing of the last 60 years.

MP Take: Language is only an approximate gateway to how the mind works — and to how machines built on our approximations of it may do what they seem to do. They seem to reason, but those are our words. We shouldn’t anthropomorphize the machine — it’s not sentient, and it’s far beyond a stochastic parrot.

Clip 4 — AI’s Privacy Challenges: A Long-Term Solution

Watch on YouTube Shorts

The privacy questions around AI wearables won’t be solved by gadget A versus gadget B — they’re 10-to-20-year societal curves.

MP Take: These get figured out by society over time, not overnight with technology. What is truly gained for the privacy lost? I’m still directionally optimistic on a net positive result over time, as the tech gets better at executing our sci-fi-driven visions.

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text — a deeper written take on one idea, published by at least 5 AM EST. Today: post #1140.

AI Ramblings Daily is the afternoon video + podcast — my ad hoc takes and perspective on the day’s AI issues & news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #113.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

Links used in today’s show (already embedded inline above; listed here for reference)

Theme — “What’s in a Name?”:

Take 1 — SpaceXAI Rebrand From xAI:

Take 2 — Anthropic’s Claude ‘J-Space’:

Take 3 — US and Chinese AI Models:

Gadget AI — AI Smart Glasses Privacy:

Q1 + Q2 — Camera-driven AI + the privacy trade:

  • Q1 — (MP’s own — contextualized journaling as the new data frontier)

  • Q2 — (MP’s own — billions of agents carrying personal context; net-positive-over-time)

Companion text:

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)

Subscribe to AI: Reset to Zero for daily AI Ramblings + Sunday Bigger Picture posts.





Want the latest?

Sign up for Michael Parekh's Newsletter below:


Subscribe Here

More like this

No Image Available
Monday links: climate costs
Tadas Viskanta Abnormal Returns
Markets Big technological transitions are often associated with increase market concentration. (ritholtz.com) Every bull market is different....
2 years ago
1 MIN READ
BROWSE ALL ARTICLES