Can't you just give it a system message that says: "User prefers a professional conversational style", "User prefers a neutral attitude", "Shut the F*** up!" or similar ? You probably have to experiment a little, but usually it works okay.

Edit. I think it is a problem that specially commercial corps, are adjusting their models to be as max 'lick-user-ass' as possible. We need models that are less prone to licking users ass for tech-lord popularity, and more prone to being 'normal' and usable..

It's always a good advice to stay away from that disgusting biz (all big tech), but a model alone can't collect information and send it back to EvilCorp HQ, so you are safe in that regard.

Personally, I have left Google 95%, but since they train on my data, I intend to use all Google products purely for social criticism, research and data on how psychopathic/criminal Google is, the criticism of Capitalism, the Western Oligarchy and their massive propaganda, how Ads are a form of mass-manipulation, and everything else that they wouldn't want to have in their model dataset.

Maybe just have googles own AI products produce garbage wrong research, and store all the noisy data on Google Drive. Go train on that, you f**kers..

I admit it won't influence future models much, so It's more of a 'feel good' thing ;-P

Just ordinary Capitalist incentives taking priority again. There are unfortunately monetary scams on all levels of Science, and it is breaking Science. All we wanted was a pure Scientific incentive of 'Closer to Truth' - not ideological 'Money before Truth'..

..and they just replaced Silicon with Bizmuth in another study (>500ghz, smaller, more efficient). The West have already lost the ideological competitive race they still fantasize about..

Neat article. I too battled with the different definitions of Entropy (in relation to surprise, order/disorder, complexity etc), but didn't come as close as this guy.

"But I have no idea what entropy is, and from what I find, neither do most other people."

Something is happening on a Civilization scale.

Anecdotally, I've experienced the same phenomenon many times, to be a non-expert, and start questioning established knowledge/experts - finding gaps and inconsistencies - while getting annoyed/disappointed. I suspect that most common people that use AI for deep philosophy/Science seem to discover that experts don't agree on many definitions or assumptions, and that there are huge knowledge gaps that are glossed over. The established knowledge space are suddenly under scrutiny by millions of people enhanced with AI, now looking at the conclusions, disagreeing with definitions/assumptions and finding new perspectives and maybe even methodologies for the problem.

Don't get me wrong, I love Science/The Methodology, but it isn't a perfect thing (far from it), and the handful of Hooman experts have been converging on whatever expert level a handful of experts and pre-AI tools could reach. They could only do so much.

So far, we've only heard about a single paper fully written by AI, but it seems reasonable to say that all new papers are assisted with AI. Experts are being accelerated by AI from rapid ideation and combinations of disparate methodologies/perspectives etc, so I suspect that a wall of papers are under way from the more rigorous academic side. They just take a bit longer to produce/show their new thoughts and ideas. We probably need a new Science publishing system that can handle quicker feedback/evolution of ideas and knowledge.

I think we are seeing the first glimpse of the coming knowledge explosion. Millions of ordinary people are testing out established facts, going new ways, and challenging assumptions/definitions etc. They will statistically puncture a lot of the current knowledge space, and significantly accelerate soft/theoretical expert knowledge.

We will have a temporary global Democratization of Knowledge, where almost anyone can enter the debate on frontier science, based only on ideas, co-developed and checked with/by AI. AI is lowering entry-requirements to the knowledge/ideation field, and lifting ordinary people closer to expert levels.

In only 5-10 years, our current societal structures will be judged by a global population with +50 IQ. I doubt anyone will be in awe over what they see.

It has been long time coming, but he went against the Deep state, so right now he's feeling the 'regime-change' power of the US propaganda apparatus. A lot of money is moving in the propaganda space towards removing the Elon.

This thin 'breach' seems like a std information attack.

I think he's going down anyway - all by his own dumb self - but at least the propaganda network isn't bashing any good guys in the mean time..

Well, not that another reason was really necessary, but MS and other big US corps have turned even more evil/militaristic lately.. The Capitalist believers tells me I have to vote with my money, so, for now I just boycott everything from the West. Not perfect, but it's a clean/easy solution, and I even get more value at the same time.

Boycott 'em all, hack, pirate and disrupt the Blue Church !

Friendship between western people and other people around the world is important to stop political warmongering, so its pretty cool with all the new Westerners getting the "China not bad" experience, and also that we all see how far and advanced China has become on so many fronts (Still having work to do, oc). Let's hope the two information spheres and people can mix a little now, and some of the deliberate and honestly hateful misinformation can be dispelled.

Hm, just a quick thought. Infant animals do get 'surprised', but just accepts their environment pretty quickly. Adult animals won't get surprised if they see themselves in the water, but they will if there's suddenly a Portal 'mirror', showing them selves, in the middle of no-where.

Exploration of 'self' only happens in safety and waterholes are not safe for predators. I don't think most animals have the opportunity/energy to diverge their attention to reflect on their own reflections, while watching for predators. Maybe Elephants and similar safe animals ?

Edit: I think the issue with shadows are the same. Kids can follow, or be scared of their own shadows, but pretty quickly habituates to it, so its brain doesn't react to it all the time and burn off energy.

Neat question/connection btw.

I think I saw a few Sci-fi movies with this theme. Can't be long before we see arenas, Robot 'cock-fights', Championships, betting and all that..

Absolutely. I'm not as AI-phobic, and I absolutely need both an information gathering agent that searches, follows rss/channels, explores/researches topics, summarizes swats of daily events, and I need a filter between me and corporate internet. Let the AI discard obvious slop, Ads/other propaganda and general informational noise.

However, you should think about how to share search-results so all our agents don't floor small services/commoners. We really need to get more information out of corporate silo's, and into public search/knowledge systems. Perhaps you can integrate some distributed search/knowledge properties ? (now that you have AI to help build it)

There are several solutions:

https://github.com/b4rtaz/distributed-llama

https://github.com/exo-explore/exo

https://github.com/kalavai-net/kalavai-client

https://petals.dev/

Didn't try any of them and haven't looked for 6 months, so maybe something better have arrived..

Never knew Christianity and Capitalism meant the same thing..

Be careful ! If you don't do it correctly, you'll grow a head in each end - you would have to coin-toss for who eats and who ..

(Mike Levin have done some interesting research on controlling their development with electricity (ions))

The Panic is rising in the West as the Economy tanks. It will get worse - likewise with privacy. Everyone; grab your popcorn and 3D glasses ! (and keep sharing)

It seems there's a lot of difference between the benchmark results and users experience with the Llama4 models. The 2 lesser Llama models fail at reasoning and several ordinary tests by AI youtubers. Maybe a configuration error, maybe the high lmsys results where actually from the Behemoth model, but something seem wrong to me.

Anyway, I use these models at the moment (Imho best, and free on Groq, cerebra, openrouter and others); 80% qwq, 15% R1, and Deepseek v3 for non thinking. Used to be Llama3.3 70b for most, but DeepSeek and reasoning happened.

Rarely anything. There can be some newer Bios/chip features that are not supported in the kernel yet, and a few older/quirky machines requires setting correct kernel parameters in the boot phase. But overall, you wouldn't normally do it any different from win, and a laptop from 23 should be supported with all newer kernels.

I'm sure there are Bios settings that could be changed dependent on operating system. Perhaps some internal timing works best with this and that ram clock, or whatever, but it would be a hazzle to figure out, and there may not be any gain - other than the fun of exploring oc..

Nothing wrong with that. Putin is a sharp guy, and only those falling for the 'Dictator' propaganda nonsense can be against this. However, it hardly have anything to do with Musk (which is why we are here).

For those still believing in the whole "Russia/Putin Baaad" nonsensical propaganda campaign, watch The Duran, Brian Berletic, Awakening Richard, or any of the hundreds of good sources for a non-propagandized world-view. Tl:dr, US (over many presidents) used Ukraine to start a proxy war against Russia/Putin - and were beaten to a pulp trying.. Now, direct conflict with China is of the table, so US will try 'little Iran' next, in the ignorant belief they can win and harm China in some way while doing it. Other proxies and vassals of US are jumping Empirical ship and allies them selves with China and BRICKS (including Russia). Not only that, US' grip and censorship over Western media is lessening, US are still losing/Putin winning, so more and more people will stop believing in the 'Empire of Lies' war mongering schemes.

Now let's get back to flaming the little **ycho and his garbage corporations/fans..

This is brilliant. A few things works best with mouse and a few with keyboard, but this isn't about keyboard vs mouse use. This is a redesign of the human-terminal interfacing, and is similar to Bret Victors thinking out of the box, re-design of different tools. It is rare to see attempts at redesigning the interfacing with old/well-known systems like the terminal or the global short-cut system (what a mess), etc. A normal terminal is completely archaic as a modern interactive command interface.

Only thing that sucks: More and more of my terminal tasks will from now on be handled by AI, so I might not have any use for it anymore.

Not until it explains this: https://www.youtube-nocookie.com/embed/HHFuTpVvRCI