An artificial intelligence can be anything but one thing: to be a soul… or?
Awesome this PS 4 game!
Detroit: Become Human - Shorts: Chloe | PS4 (youtube.com)
What Happens if You Don’t Let Chloe Go - Detroit: Become Human Menu Keep Chloe (youtube.com)
Im curious if its possible to connect tensorflowjs with babylonjs, for example to replace or axcelerate shader calculations.
Lets say we want to create a fluid simulation, so we can implement one of the exising approaches (example [2002.09405] Learning to Simulate Complex Physics with Graph Networks). Btw tensorflowjs has webgpu support also.
It is possible but to be highly efficient it is a pain cause you would need to share the context.
For fluid sim particularly, an analytical approach might be faster.
AI researchers found that widely used safety training techniques failed to remove malicious behavior from large language models — and one technique even backfired, teaching the AI to recognize its triggers and better hide its bad behavior from the researchers.
Midjourney 6 is able to produce quite good results if you have enough time to write the right prompt.
Good weekend everyone,
It’s the weekend and therfor, another good time for me to share some thoughts and some ‘philosophy’ (as I did previously in this forum) 
I suppose this thread is the one to post about my latest experience with the AI. Isn’t it?
As I’m about to release the next version of my project ‘MOE Space Museum’ with the adding of an ‘Auditorium’ featuring a selection of ~70 great works of Music from Medieval to Atomic Era selected by the AI (in this case, chatGPT), I find it interesting, as a sidenote, to share my thoughts and experience from this collaboration (with both chatGPT and MS bing AIs).
In particular, with this collaboration, one thing that striked me and is really starting to worry me is 'How much these (western-originated yet pretendingly ‘global’) AIs are ‘oriented’.
Despite from my multiple prompts/rules to make “the most ‘diverse’, ‘objective’ and ‘global’ selection”, the results I get are obviously very much orientated towards just ‘western/occidental’ culture. In a way, I believe it’s understandable, since based on data… and since most of the feed comes from the ‘mass’ western-culture orientated users. So, this is where I’d like to alarm people (people from the world) on what’s about to happen… and more importantly, how we can still change/influence things for the future. Of course, in my opinion only:
I believe the young/child AI from the 2020’s is now growing into a teenager. Despite it doesn’'t have all the ‘emotional issues’ of a human teenager 
it is still looking for it’s way to build up, grow as an adult. Again, in my opinion - it’s still gonna take a number more of data and interactions (plus of course, work on the algorythm and more/better rules)…before the AI will eventually (likely) become a true partner in our lifes and businesses.
I believe I stated this before but I’m gonna say it again: “We ALL have a responsibility” in how we make this thingy grow. We need to ‘nurture’ and ‘educate’ it. And this, in a way that’s not just matching our own limited and selfish lifes, but accounts ‘humanity’ altogether.
While building my virtual museum selected from the AI, I kept repeating to the AI that “this isn’t about ‘my preferences’”. It acknowledges it but, persists in making orientated ‘choices’ (or, in other words, giving what I believe to be ‘orientated’ responses). Though, eventually, when I start ‘challenging it’, these answers show something very different from the initial!
My thought is that there should be more of us ‘challenging the AI’, disrupting its logic and sharing different input/data. If we want an AI that’s not just a tool (or a weapon) to claim cultural hegemony for a certain type of governance/ideology, … well, we still have a long way to go. Starting with becoming better humans, more open, more doubtful and more empathetic. Because in the end, the AI is not just a machine; It’s essentially a machine made by…humans! 
Every child with an empty mind adopts the observations and actions of its environment unfiltered. It does not know what it means to act “right”. They simply do the things that the people around them do. Who among the adults actually knows the meaning of “right”? A word that can be interpreted individually by everyone.
I’ll go one step further: most adults don’t know what they are doing because their acquired beliefs have become second nature and are no longer under their control.
The AIs we breed will always reflect who we are on average.
It will be a very long time before we become a highly cultured free altruistic civilization.
The road is long, but we have plenty of time.
We are not limited to a single life!
*“Wanting to be sensible in a crazy world is a madness in itself.” *
– René Descartes
While the sensible ones want to develop a decent AI,
the stupid ones are already storming the castle 
A fine way to illustrate the problem. We are basically still just barbarians with a smartphone (although we like to think that we are better than our ancestors).
I hate to leave you standing on this extreme point of view.
I like to use the example of a battery. Energy is created by the difference between two different poles. If you add up both sides, you get zero. No energy. No life.
There must be differences. Chaos and order. Health and illness. Cause and effect. We are prisoners of causality. How can freedom be defined if there are no obstacles?
Oh, sorry. Misunderstanding here. That’s out of its context and… yes, may sound extreme as such. But I’m nothing desperate nor having extreme point of view. It’s actually all the contrary. I’d like people to have a much less extreme and polarized point of view than the tendancy we can sense these days.
Completely agree 
…And then, I can consider myself very lucky. Most if not all of the people I connect with or need to interact with are ‘good’ people. Starting with all the ones in this forum. I’m also not in ‘survival mode’, …see, I can even find time to make some ‘pub level’ philosophy While, most people out there cannot afford this luxury. Reason why I sometimes have the feeling that, in my condition and within my reach, I should also try to ‘give back’ a little (at least I’m trying to ). Have a great day 
Do not define. Then, you are truely free.
Just wanted to share my latest experience with the new chatGPT (v4) …and thus, explaining once again why I say “Don’t let the AI take over” or “don’t trust it” (put it the way you want).
Like, this morning, I went back to the AI to add a new category for my ‘Space Museum’. As a reminder, the space museum features great works and achievements that showcase ‘human evolution’ - seen through the AI. All selection and description come from the AI (bing and chatGPT).
So, the next category I’m adding is ‘Great Rulers’. Throughout history, a diverse and global selection of 60 of the greatest rulers. I asked chatGPT (v4) to pull me a list. I asked for a diverse selection, based on achievements, governance models and impact on transforming changes in society/humanity. I also asked to order it, stating the Era, Governance Model or alliegiance.
So, it pulled me a first list. And the list was pretty good and diverse. So, I asked it to build on it, pulling the description and filling-in the information (of era/period in history and governance).
And that’s where it started to ‘bug’. First, it didn’t remember of the original list. Despite for stating that it will (quote) “create all descriptions for the list”. The list it just pulled 30sec. ago. Instead, it changed both the order and (even worth) the list itself. Ignoring some leaders and adding others. And just to make this clear, I’m not really a ‘newbie’ collaborating with the AI. I know of this kind of jokes, so I always make the clear prompts when referring to a discussion (at least I think so 
)
Anyway, that’s one thing. But the other is a lot more disturbing. It gave me ‘wrong’ information. I mean I’m not an expert in the history of civilization. But what I got for the description of the ‘Era’ is clearly wrong.
For Era/Period in history, chatGPT v4 did the following (i.a):
For ‘Suleiman’ (1520-1566), for ‘Philip II’ ( 1556-1598), for ‘Elizabeth I’ (1558-1603)… and half a dozen of others in the 15th, 16th or 17th century, the AI stated the Era is ‘Early Modern’. Well, there’s no faen chance that this would be ‘Early Modern’ in any civ. To enter ‘Early Modern’ era, one would first need to perform an ‘Industrial’ revolution. And clearly, this didn’t happen anywhere in the world at these date in history
My prompt was clear: A selection from ‘Ancient’ to ‘Atomic’ Eras. So, how come the AI includes rulers from after 1980 to present?
Anyway, what I guess I wanted to say and share here is two things. First, don’t think the AI will make just all the work for you. And second, don’t lead the AI (same as don’t lead other ‘human’ collaborators) if you have no faen clue what you are talking about. The AI might be smart and of great help, but it still needs you for the lead. Of course, as always, my opinion only
Have you tried the pi.ai?
yeah AI is not as smart as i orginally thought it to be. It is incredible and it will get better but currently no… ( also tested on th new 4o model )
In my test situation I was looking for a code solution to something. While it had ideas around the topic , it could not produce production ready code at all, well not unless you as a programmer basically spelt out in detail - exactly what it needs to create for you and when I say in detail , im not meaning just some detail , im meaning you have to instruct it with all the knowledge of the code you want, for it to give you the code you want, which defeats the purpose of trying to use it to give you something you dont know , as in - be smart on your behalf.
this was in relation to using multiple streams ( read , duplex/transform , write ) and piping in nodejs. I know why it struggled… because I also struggled to find the correct information using the internet. not even nodejs documentation gives good examples on how to use eveything properly in different scenarios , and 99% of what you get from google are all 1 million developers personal blog posts / medium page posts … that copy pasted the documentation code , or 2 or 3 other snippets that hardly do much or actually have bad code etc.
