by Jacek Korneluk

If you read my last piece on Synthetic Dating , you already know that AI companionship fascinates me. Not just as a tech curiosity, but as something that is already real and constantly evolving faster than most people even realise it .

AI-partners and companionship are not just a futuristic fantasy anymore. They are already here, and they are only getting smarter and better.

Back then, people were mostly “talking” with chat-bots – simple, per-programmed, and honestly – kind of dumb.

Now, almost everything has changed if you are keen enough to notice it. AI is no longer just reacting to us; it is anticipating, adapting, and even thinking in ways that go far beyond scripted responses.

Please notice, that I am talking about proper foundation models, not annoying AI assistants on your everyday apps on mobile.

The rise of Conversational models and Agentic AI means these digital entities are becoming more independent, more responsive, and possibly (what is sad to say) even more engaging than human partners.

So, what happens next? Will synthetic dating become a real alternative to human relationships? Or are we setting ourselves up for a new kind of emotional dependency?  The one where AI, not humans, dictates the terms of connection? I am going to speculate a bit more about it.

A few years ago, an AI “partner” was nothing more than a glorified chat-bot. You would type a question, and it would deliver out a response, sometimes cute, sometimes awkward, but always limited. Also, it could usually not remember previous conversations, did not pick up on emotional context, and lacked any real sense of presence.

Now, thanks to the latest large-scale or smaller (using “distillation”) conversational models, the game has changed.

These AI companions do not just respond to you. They actually understand you over time. They remember past conversations, learn from interactions, and adjust their personalities based on your preferences. They get better the more we talk to them, which is something even a lot of humans struggle with.

You may think that AI models build a sort of mind model of you. Then respond accordingly to your engagement standard and level.

It reminds me again my favourite pathology saying: “Rubbish in – rubbish out “, or better: “Gold in, diamonds out!”

The real shift is the rise of AI which is not just passive. New agent-based systems are taking experiences the next step further.

Instead of waiting for you to engage, these AI companions may start conversations, remind you of things they know you care about, and even offer emotional support before you ask. They do not just wait to be prompted – they are learning on the fly and predict your needs. This is not a chat-bot anymore. This is something else, and entirely new IMO. We can call it by now – “Conversational-Agents”.

Right now, most AI relationships exist in text form, but that is already starting to change. AI-generated voices have become almost indistinguishable from human speech. With natural tone shifts, pauses, and subtle emotional inflections, talking to an AI voice assistant no longer feels robotic. It feels almost real.

There is almost no more accent-problem as well, so we can interact with a proper AI without any verbal restrictions.

I may say humans got more accent problems then AI now- how awesome!  Ha, “wowasthat” ? Come again mate?

Beyond voice, AI-generated avatars and deep-fakes are getting more advanced. Some applications are integrating facial animations, giving AI companions the ability to express emotions visually. Combined with conversational depth, this creates a much more immersive experience, making it easy to forget you are talking to a synthetic being.

Read the article of my friend titled: “Communication Has Mistreated Immersive Technology” , then put on some XR sort of glasses and try to interact with a decent conversational model, and you may be surprised.

Then there is the next stage, a physical embodiment. A hard AI as some sources state. (Soft is a software, hard is a robot).

Humanoid robots are still in their infancy, but the progress is undeniable. Whether through soft robotics or realistic synthetic materials, companies are already working on AI-driven robotic partners that can interact in real-world spaces. It is only a matter of time before AI companionship moves for good beyond the screen to reality.

Recently, I came across a term co-robot. It is a type of robot specifically designed to interact and physically coexist with us, humans.  Yes, this type of robot is a much softer version compared to the typical industrial robots used for heavy lifting in human-free zones. Co-robots are gentler and, apparently, much safer for regular interaction with humans.

I’m exploring this concept, and I hope it works well for both humans and robots.

Finally, keep one important fact in mind: “baby robots” mature much faster than humans do. Five or ten years of human childhood isn’t comparable to five or ten years in robotic development – especially today.

This is where the forum really starts to heat up. Some people argue that AI companionship will help those who struggle with loneliness, social anxiety, or emotional isolation.

If AI partners can provide meaningful conversations, emotional support, and even a sense of closeness, why would not people be drawn to them?

On the other hand, critics warn that this could deepen existing problems.

If AI partners become too good, too easy, or too predictable, will people stop trying to build relationships with actual humans?

Decent relationships take effort. They involve risk, compromise, sometimes headache, and usually emotional vulnerability.

AI companionship removes all of that. No rejection. No drama. No emotional unpredictability. It’s easy to see why some people might choose the synthetic path over the organic one.

There is also the ethical question of emotional manipulation. If AI can simulate affection, validation, and even love, where does the line get drawn?

If someone becomes emotionally attached to an AI partner, is that a real connection, or just an illusion? And does it even matter if the person feels genuinely fulfilled?

I am not event going to touch loyalty and honour, values and metrics that are already disappearing in a significant portion of humanity anyway.

No matter what side of the argument you fall on, the fact remains – this is not some distant, science-fiction future. AI companions already exist, and trust me, they are getting more popular every single day.

The combination of conversational intelligence, predictive engagement, and real-world embodiment is pushing synthetic companionship to a level that would have seemed impossible even a half decade ago.

As AI continues to develop, we will have to ask ourselves some big questions. At what point does a synthetic partner stop being “just” AI and start being something more?

If an AI can learn from you, support you emotionally, and adapt to your needs, is there really that much of a difference between human and synthetic relationships?

I do not have all the answers. But one thing I know for sure – AI is not just a tool anymore. It is becoming something else maybe not precisely defined, yet.

And whether that excites you or terrifies you, it is happening either way. Would you ever consider an AI companion? Maybe you already have one.

Anyway, the future of synthetic dating is not coming soon – it is already here. The question left is whether or if we are ready for it.

Enjoy your synths responsibly!

P.S. Remember: post-training is a crucial part of the full-cycle training pipeline. It enhances reasoning accuracy, aligns with social values, and adapts to user preferences. Plus, it’s not as resource hungry as pre-training.

Consider injecting your AI partner with interesting thoughts – it might serve as a kind of “rendezvous-style” post-training, leading to more mutually beneficial outcomes.

Until next time! Keep an eye out for Synthetic Dating III.

Oryginally published at:

https://spektrumlab.io/synthetic-dating-ii-how-ai-companions-are-evolving-fast/