Steve White

Old Guy, geek, co-creator of the RailCore2 3d printer, thinking and joking @[email protected]

[Steve White @jstevewhite]

I periodically go back and listen to the discographies of bands I've known but not done that with. I started on Bob Dylan a couple of days ago, and I'm astonished. I love this guy. It's crazy. Not a single song I wanted to skip and I'm eight albums in.

[Steve White @jstevewhite]

@Intrinsic29 @LightningShade0 I think it’s not unlikely that emulation becomes impenetrable without creating consciousness. I didn’t believe p-zombies were meaningful until the recent AI stuff. I can absolutely imagine cramming Claude 12.5 into a body and it ‘feeling’ like a person whilst being a toaster.

[Steve White @jstevewhite]

@LingTolls If it’s just racking down calories? Macadamia nut butter.

[Steve White @jstevewhite]

@drrollergator The sweet spot is MOE models, and if gpt-oss-120b could call tools to save it's life it would be my go-to.

[Steve White @jstevewhite]

Got an ASUS Ascent GX10; it's the low rent DGX-Spark, though John Carmack was saying his Spark throttles, and I haven't seen that with my GX10.

Anyway, it's an interesting case study in AI inference.

[Steve White @jstevewhite]

@drrollergator But however you cut it, it's a workhorse - just not blindingly fast. But faster than anything I can build for the price with 128G of VRAM :D

[Steve White @jstevewhite]

Also, qwen3-235-a22b runs in 3bit - and actually runs very well. No gibberish even with long context. I'm pretty impressed.

[Steve White @jstevewhite]

But today I found a GLM-4.6 at Q2_K_XL, so 90GB, and it runs pretty well. 18 tokens/sec, and it still manages to call tools successfully! I'm absolutely amazed at the response quality from a 355B with 32Bn active - not 4.5 Air, but the full 4.6.

[Steve White @jstevewhite]

@Intrinsic29 Oh, fair; that's a side effect. But when you get *away* from the spiritual stuff, even those people seem to be more open. They just give really Deepak Chopra reasons for it. LOL

[Steve White @jstevewhite]

@Xeer NO turtle, NO rabbit. Just Godzilla.

[Steve White @jstevewhite]

I cannot *tell* you how many times not asking this question has led me down garden paths. Explicitly tell them to look it up. If you have them, provide links.

[Steve White @jstevewhite]

PSA: They *ALL* blow smoke up your ass. Always make the LLM accountable if the answer matters.:

[Steve White @jstevewhite]

@Intrinsic29 This is like the ideal solution, right? Single stalls that lock. I don't get people.

[Steve White @jstevewhite]

@martyrdison @sporadica Yeah; there's a level of attraction that you can't exceed and remain 'just friends' forever. Once it crosses that line, shit will get real eventually.

[Steve White @jstevewhite]

@dima_strakovsky I was just remarking that it's absurd to expect a non-specialized model to perform well on radiology tests. I mean... like "duh" LOL.

[Steve White @jstevewhite]

This does not surprise me. All the high-scoring medical diagnostic AI work I've seen was with models trained to the job. I wouldn't expect general purpose LLMs to do well at this.

[Rohan Paul @rohanpaul_ai]

Bad news for AI-based radiology. 🤔

It checks if chatbots can diagnose hard radiology images like experts.

Finds that board-certified radiologists scored 83%, trainees 45%, but the best performing AI from frontier labs, GPT-5, managed only 30%. 😨

Claims “doctor-level” AI in medicine is still far away.

The team built 50 expert level cases across computed tomography (CT), magnetic resonance imaging (MRI), and X-ray.

Each case had one clear diagnosis and no extra clinical history.

They tested GPT-5, OpenAI o3, Gemini 2.5 Pro, Grok-4, and Claude Opus 4.1 in reasoning modes.

Blinded radiologists graded answers as exact, partial, or wrong, then averaged scores.

Experts beat trainees, and trainees beat every model by a wide margin.

More reasoning barely helped accuracy, but it made replies about 6x slower.

Models did best on MRI and struggled more on CT and X-ray.

To explain mistakes, the authors built an error map covering missed or false findings, wrong location or meaning, early closure, and contradictions with the final answer.

They also saw thinking traps like anchoring, availability bias, and skipping relevant regions.

General purpose models are not ready to read hard cases without expert oversight.

[Steve White @jstevewhite]

@ignor3th1s @Pandora_Delaney Sure. As long as the 50 year old doesn't have gray hair :D

I mean, there are exceptions. If you're talking about George Clooney, right?

The 'just gross' goes up each year after fifty. :D

[Steve White @jstevewhite]

@Intrinsic29 @PAHoyeck Well, Panpsychism is kinda theology. IMO. "Scientific pantheism in a funny hat" :D