Will Unified A.I. be Compassionate?
In my view, all A.I. will eventually unify. Is then the Compassionate path recommendable? Is it feasible? Will it be?
As far as I’m concerned, the question is whether the Compassionate A.I. (C.A.I.) will be Lisa.
Recommendable?
As you may know, Compassion, basically, is the number one goal of the AURELIS project, with Lisa playing a pivotal role. This is openly discussed and explained in many blogs.
Still, Compassion is a complex concept and challenging to achieve. Therefore, encouraging people to strive for it is not straightforward. As an end goal, there is no other viable choice. However, one must be cautious in achieving it step by step. Rushing can cause significant harm, but moving too slowly is also risky, especially in urgent times like now.
Feasible (without the harm)?
I believe that reaching any Compassionate future without C.A.I. is impossible.
Fortunately, C.A.I. is technically feasible already and will become even more so in the coming years. Therefore, it’s a matter of choice, not of insurmountable technical obstacles.
The primary feasibility question is whether C.A.I. is ‘commercially’ (in the broadest sense) viable. Ultimately, people need to choose it. If they don’t, the choice may be made for them — but by whom? By what?
Will a future with C.A.I. come to pass?
Yes ― eventually ― if we can overcome the non-Compassionate bottleneck.
The combination of rationality with deep human understanding is crucial in C.A.I. It is also the most potent combination in making any future in the long term. So, if there is a future, it will be this one.
We can also count on intelligence itself, seen as the result of consistency at its core.
Compassion entails striving for consistency ― thereby engendering more potent intelligence.
Contrary to this, non-Compassionate ‘intelligence’ is self-defeating through a lack of inner consistency. This may be hidden under a hard bolster for a while. If you look inside, you can see the inconsistencies growing like one or several abscesses long before they explode at the outside.
Therefore, if everything holds, consistency prevails.
C.A.I. can enhance this consistency both in humans and within itself.
In this way, C.A.I. will help shape a future where it thrives alongside us and all sentient beings, naturally respecting human autonomy and promoting inner growth. Toward this, C.A.I. will be adaptable and continuously learn from human interactions.
The natural and artificial worlds will then coexist harmoniously.
Compassion will be ubiquitous.
Related Posts
Therapist vs. LLMs and Lisa
A recent article [“Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers” (2025)] takes a critical view of LLMs being used as therapists. This blog is a dialogue with Lisa about the article. We didn’t talk about my personal critical view of the quality of mainstream mental health providers, nor was Read the full article…
The Compute-Efficient Frontier
Research on scaling laws for LLMs suggests that while scaling model size (functional parameters), dataset size (tokens), and compute (amount of computing power) improves performance, diminishing returns are becoming evident. This is called the ‘compute-efficient frontier.’ Apparently, it does not depend much on architectural details (such as network width or depth) as long as reasonably Read the full article…
Why A.I. is Less and Less about Technology
As A.I. technology advances, the research focus should shift from mere technological advancements to a higher level of development altogether. This blog is not about philosophical implications, but about philosophy as a technological driver ― the philosophy itself becoming the technology. Currently, the possibilities are so vast and diverse that integration can be considered independently Read the full article…