Enigma Spectre

*archaic croaky voice* “I remember when AI was just called predictive analytics” *fades to dust*

[Enigma Spectre @echo_fractal]

Although I think quote is better as the meme text on the ancient aliens guy….

[Enigma Spectre @echo_fractal]

Y’all are ready for quantum, right?

Error correction is low hanging fruit rn, and it’s gonna go fast

[Google Quantum AI @GoogleQuantumAI]

Published today in @Nature: We've unified calibration with computation on our Willow processor, training a reinforcement learning agent to stabilize the logical qubit and pave the way towards a quantum computer that continuously learns from its errors.
Read more: https://t.co/MUAaxkfimN

[Enigma Spectre @echo_fractal]

Here what I was working with and some of the results I got. I didn't really mean to nerdsnipe myself with this today though, so I am going to leave it here for now. LMK if anyone thinks this is interesting and maybe i'll come back to it.

Cheers.

https://t.co/HV4xjrumbT

[Enigma Spectre @echo_fractal]

Has anyone tried probing the j-space with “the same prompt” in various different languages (assuming the model is multi-lingual of course)?

Or by asking for output in a different language from prompt?

[Enigma Spectre @echo_fractal]

so, I tried to run a quick proof of concept and here are the first results. the prompt is "The capital of France is", and the target word here is "Paris" (both translated as needed). Qwen 0.5B for simplicity rn.

Interesting results TBH, mid-layers only

[Enigma Spectre @echo_fractal]

How your email finds me

[Enigma Spectre @echo_fractal]

@bebeal Hard to fit in char limit, but basically a simulated manifold with a special black hole. It acts like a diffusion RNN. “Tokens” perturb a field, time steps allow waves to prop, senosors measure state, next token predicted. Plus memory and learned init conditions make a model

[Enigma Spectre @echo_fractal]

Another solid “wrapped” for the year

[Enigma Spectre @echo_fractal]

@bebeal Tbh the project in this thread been sitting idle a while on my end, and idk if I have a repo public.

This info has been super helpful though, and I’ll reply with a link later when I have one.

[Enigma Spectre @echo_fractal]

@bebeal Ty for sharing this

[Enigma Spectre @echo_fractal]

To be honest I had kinda missed the Landauer erasure cost in my own work, and there are a couple other really neat things here.

https://t.co/XVhzmPRu2r

[bebeal @bebeal]

Project Page: https://t.co/gFsMHbqgmj
Paper: https://t.co/BTNgUx97LV

[Enigma Spectre @echo_fractal]

*whistles and motions nonchalantly af*

https://t.co/iEGLc8L0Js

[Unconventional AI @unconvAI]

Introducing our first model, Un-0!

We trained an image generator powered by a backbone of coupled oscillators in place of a more traditional conventional neural network.

[Enigma Spectre @echo_fractal]

Me in the model’s J-space:

“J’accuse, and the gradients agree.”

[Anthropic @AnthropicAI]

New Anthropic research: A global workspace in language models.

Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.

We found a strikingly similar divide inside Claude.

[Enigma Spectre @echo_fractal]

I’m actually taken aback by this and it didn’t click when I saw the announcements earlier.

Someone, please be excited with me. Folks that ik irl won’t even know the meme 😭

[Enigma Spectre @echo_fractal]

Wait… THEY DIDNT JUST BOIL WATER INTO STEAM?

Plasma windmill go brrrrr

[𝓜𝒂𝒖𝒓𝒊𝔃𝒊𝒐 𝗜𝒃𝛼 @Dragonmaurizio]

⚡️ Fusion technical milestone.

Wisconsin based Startup company Realta Fusion has successfully converted plasma energy into directly usable electricity without the use of cooling hardware and steam turbines connected to a generator.
The demonstration of Direct Energy Conversion (DEC) was achieved with a device called Wisconsin HTS Axisymmetric Mirror which includes the use of High Temperature Superconducting magnets (HTS).

The prototype uses a magnetic mirror configuration to slow down charged 𝛼-particles at the end of the reactor, capturing their kinetic energy and converting them into electrical power in a single stage.
The converter is composed of three finely meshed grids; an electrically grounded grid, an electron repulsion grid, and an ion collector grid.
The unprecedented trial drew multiple amps to illuminate an array of lightbulbs.
@RealtaFusion credits.

🔗 https://t.co/e6ZW8kCXQ6

[Enigma Spectre @echo_fractal]

Update: it was, and still is, delicious.

Cheers

[Enigma Spectre @echo_fractal]

I have acquired a good brisket, and the smoking will commence tonight.

It’s a bit of a tradition for me around this time of year, but it took a while to find the right cut in my new hometown. Glad I found the shop that I did though. Cheers

[Enigma Spectre @echo_fractal]

@neura_eth 🤩
Those are beautiful, and my two favorite strains.

[Enigma Spectre @echo_fractal]

It was a beautiful sunrise here today.

Hope that you all have a wonderful day out there. Cheers

[Enigma Spectre @echo_fractal]

@jhouse678 @mathelirium I’ve wondered if there’s some form of anti-hawking radiation to resolve the paradox.

[Enigma Spectre @echo_fractal]

Embarrassing that I did not think of this tbh, given my interests in both mycology and reservoir compute.

Anyways, that’s kinda neat. Cheers

[Eben Bayer @ebenbayer]

Mycelium Chips?

No, not the kind you eat. the kind funded by DARPA that do inference based on the grown structure of mycelium.

its a cyborg onramp

[Enigma Spectre @echo_fractal]

lol

https://t.co/jcTJ1BWbFq

[Philo Groves @PhiloGroves]

This is a magnetoencephalography (MEG) machine. Meta trained brain2qwerty v2 by having 9 volunteers wear one while typing for 10 hours each.

[Enigma Spectre @echo_fractal]

Certainly, Meta will not use this technology to sell ads or for other more nefarious purposes.

Cheers

[AI at Meta @AIatMeta]

We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2.

Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication.

We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating.

🧵👇