Pekka Lund

Mul on S25 Ultrassa 80% lataus päällä ja siltikään ei tarvii ladata joka päivä, vaikka lataan yleensä kun putoo 30% paikkeille. Eli käytän jatkuvasti vaan puolta kapasiteettia. Ei oo siten syytä olla säästämättä akkua.

Yep. Reuters said:

"Beijing is keen to showcase progress in domestic chip self-sufficiency efforts through advances in frontier AI models, encouraging domestic firms to rely on less advanced Chinese chips for training and inference"

But failed to provide clarity on the training part.

I didn't find the press release itself.

Chinese AI startup Zhipu releases new flagship model GLM-5 China's Zhipu AI released its latest artificial intelligence model on Wednesday, joining a wave of domestic rivals unveiling more sophisticated versions of the technology ahead of the Lunar New Year f...

"The latest model was developed using domestically manufactured chips for inference, including Huawei's flagship Ascend chip and products from leading industry players such as Moore Threads, Cambricon and Kunlunxin, according to the statement."

Agreed, having those in Preview stage for months seems odd, especially when release intervals seem to be generally shortening for other labs.

I guess it's about avoiding making promises about stability when they still want to be able to tweak the model. But others use point releases for that.

If you want to think it in terms of continuous experience, it would be like a film where the narrator is always describing the previous frame, not the current one.

And the narrator doesn't need to have seen any actual footage but could be just reading the script.

I'm essentially saying that the "now" is in that reporting, not in some actual experience before it. The experiencing part, which doesn't really make any sense considering all we know about science, doesn't need to happen and isn't needed for producing those internal reports.

Principle 1: Information that comes out of a brain must have been in that brain. Perhaps we can call it a computational conservation of information. Logically, nobody can think, believe, or insist on any proposition unless that proposition is represented by specific information in the brain – and in a form that affects the systems responsible for thinking, believing, and claiming. If you believe that you have a subjective, phenomenal experience – an experience of some of the information content in your head – then that belief obeys principle 1. You think it, believe it, and claim it, because your brain contains information descriptive of it.

Reporting isn't just about reporting externally, like saying aloud what you felt, but internal reporting that enables formulating thoughts about it and so on.

As Graziano states, your brain has to have formed such information first so that it could report it.

Think of it in terms of processing in the brains, where every stage takes time. So a "direct" experience would be like:

sensory processing -> experience -> reporting

And memory recall of that experience would be like:

memory fetch -> experience -> reporting

It's about where the "now" is in that.

I now tried to ask Gemini if it can find such previous answer anywhere, and it responded:

"The short answer is: No, Goff has not given a clear, unqualified "Yes" to Sean Carroll’s claim, and you are right to suspect that he evades this because it traps him in a dilemma."

BTW, you are of course free to point me to one of those previous discussions you mentioned, or any other source, if you have provided an answer to that question elsewhere.

Arena | Benchmark & Compare the Best AI Models Chat with multiple AI models side-by-side. Compare ChatGPT, Claude, Gemini, and other top LLMs. Crowdsourced benchmarks and leaderboards.

Was it so that you can't access Nano Banana Pro at all in the Gemini App either without a paid plan? Are there places where it can be tested for free?

It's available in arena.ai as gemini-3-pro-image-preview-2k but gave only errors when I tried that same prompt there.

And it also added this to the end of its thinking:

"The absurd scale of the ants is meant to be a compelling contrast."

Clearly it tries to find a way to make an odd request meaningful, as you wouldn't see normal size ants.

You get thoughts only for Nano Banana Pro, both in AI Studio and the Gemini app. It requires paid plan in AI studio.

I tried that French prompt with Nano Banana Pro on the Gemini App and it thought:

"I'm now focusing on a realistic street scene with people interacting with giant ants"

It doesn't make any sense to claim I'm not bothered to find out what your view is when I'm here repeatedly trying to ask what your view is and you keep on refusing to answer.

Saying whether you disagree with physicists should only take a pretty quick yes/no type of answer, and that would actually clarify the situation for me, if you genuinely want to provide such clarity, as you indicate.

I just made a point why it fails. If you are trying to sell panpsychism to me, you aren't doing a good job by failing and now actively refusing to answer the simple question of for what exactly is that explanation even needed.

YouTube video by Miles Finch Professor Denzil Dexter

It reminded me of professor Denzil Dexter.

It isn't a conflation but an explicit statement that in that case those are the same. Or alternatively that you can get reports without there ever being those experiences.

It's not uncommon that people describe events, including what they felt, that never happened. I'm saying it's the same here:

If not, you are trying to explain a tiny piece of the physical world by extending that explanation to where it's not needed. And that includes all the underlying physics of our brains. And the combination problem just becomes worse, as there's no meaningful parts to combine.

Again, if you aren't evading, why didn't you just answer?

I suspect you know you have painted yourself into a corner there. If you are arguing against physicists, you are in fact trying to compete with science.

Surely the above wasn't hard enough question for needing that evasion?

I'm trying to establish where you actually see the need for that "panpsychist explanation". So are you arguing against physicists who don't see such need at least anywhere else than human consciousness, or is it just for that?

I don't see them being ambiguous since I'm staring at that "Gemini 3 Pro Preview" selection on AI Studio every day. Same with Flash, same with Nano Banana Pro. All clearly marked Preview.

I'm pretty sure they announce 3.0 general availability next. And I expect it to be a significant improvement.

3 Pro was announced 18 Nov 2025, so if GA was released now, time from first Preview to GA would be about the same as it was with 2.5 Pro.

2.5 Pro timeline was like this:

Experimental Preview (gemini-2.5-pro-exp-03-25): March 25, 2025
Public Preview (gemini-2.5-pro-preview-05-06): May 6, 2025
Public Preview (gemini-2.5-pro-preview-06-05): June 5, 2025
General Availability (GA) Release: June 17, 2025

Check the "Model Information" section:

"Status: Preview"

Gemini 3 Developer Guide  |  Gemini API  |  Google AI for Developers Learn about the new features of Gemini 3 in the Gemini API.

Which docs?

It's clearly named Preview e.g. in AI Studio model selection and in the developer guide below. They probably don't use such markings in most docs so that they don't have to update them when versions are updated.

Gemini 2.5 Pro for example had multiple Preview versions before GA.

Why? Current one is a Preview. Wouldn't it be kind of weird if non-Preview never arrived, as it has before?

I'm sending this one back.

I paid for a Gemini 3 Pro GA and that's what I have to get.

It doesn't even load.

So it also seems to work as well as many CDs.