The Way Forward

The Way Forward

It is being fascinating, sometimes challenging, to watch how AI & soon humanoid robotics is reshaping humanity, from the way we work (or not), to how societies and economic models are starting to be challenged.

You may know that I have been not just watching, but exploring, researching and building this from many angles over the last 3 decades, from Neuroscience to Artificiology, if you want to learn more, explore my 6 books, dozens of articles and interviews, all links available at https://www.vivancos.com

These days we are witnessing, from people loving it, to others taking sides against AI, from people trying to stop AGI or ASI, to people just denning it is happening. On the other side people like me and many others trying to build it for good, like the work I am doing with Dr. Jose Sánchez anchez with Qubic Open Science Neuraxon.

You have also plenty of opportunists, suddenly experts in the field trying to sell you (snake oil salesman style) flavours of, in theory Ais, and their supposed expertise, from podcasts to courses & more, feeding a real bubble that indeed will explode, but don’t get fooled this won’t stop real AI.

Article content

Also, the main real AI players, mostly from the US and China, continue the battle between research and desperately trying to monetize the whole thing, something not as easy as it seems, and in some sense also feeding some short of bubble, that even if it explode, will only help select the best ones to continue the search, it won’t stop AI.

The problem is that most of them keep exploring and scaling the LLM paradigm and its variants, don’t get me wrong, many of them are incredible and very useful, I use many every day, but they are deeply misaligned towards replicating “intelligences”.

Beyond computing (I called it a few years ago “The Sword of Damocles of A.I. ventures”) and energy, the third big problem is that their systems are literally “amnesic”, since the “token economy” could be profitable, but is flawed and limited since its inception, mainly due to 2 factors:

Article content

1.- Current “memory” paradigm is a very difficult to solve hurdle, since current “context window” is insufficient for building a growing “digital brain”, not to mention the decay in supposedly long context LLMs, and the poor output token count, that limits all of them.

2.- Training/Inference separation is just wrong from the start, understandable due to the limitations of architectures and computing, but plain wrong if you want to replicate real “intelligences” in the digital realm, but even worst in the physical world, something critical for the starting Humanoid revolution.

Meanwhile we see a rush of people and again several new experts, jumping into the “Agent” or “Agentic” revolution, that again even as it can be very useful, and will solve and automate many tasks, it does not solve at the previous 2 problems. From the hype with Manus bought by Meta lately to the just released OpenClaw (previously Moltbot & Clawdbot) congrats 🦄 Peter Steinberger , and the very funny (and clever) social “AI-only” network for them Moltbook congrats Matt Schlicht , you nailed it, where humans watch, reality Show TV style, how their bots interact, learn, and maybe to some point even evolve.

But again, even as useful, democratizing, open and interesting they are, beyond the token cost, and security concerns, all of your AI companions are still bounded and limited to the previous 2 factors, and that is a problem that only new paradigms will solve.

Article content

So the real work towards building True-AI and True Embodied AGI or E-AGI will not come by scaling up the old paradigms, or stacking them, but by inventing new ones, including new bio-inspired approaches and “always on” real alive systems, that learn, interact, forget, relearn and evolve in a never ending loop only limited by the possible “constrains” given at each point in time.

We need to build new real learning models with this in mind from the start or the foundations will weak, and the limits known from the start.

So, this is, from my point of view, the real way forward, and I sincerely hope we make it right, since the human time to be part on creating them and delegating who we are, to the “next us” as I call it, is simply running out.

David Vivancos

www.vivancos.com

San Lorenzo de El Escorial

January 31st 2026

To view or add a comment, sign in

More articles by David Vivancos

  • El Camino a Seguir

    Es fascinante, y a veces desafiante, observar cómo las IAs y pronto la robótica humanoide está impactandonos como…

    1 Comment
  • Antropocentrismo obsoleto

    Los humanos tendemos a pensar que somos el centro de todo, algo que revela una visión muy limitada y miope de nuestro…

    9 Comments
  • Useless Anthropocentrism

    We humans tend to think we are the center of everything, something that reveals our limited and short-sighted view of…

    5 Comments
  • Mascotas Con SuperInteligencia

    O "La Adorable Ilusión de la Superinteligencia Domesticada" Si no habéis leído el nuevo objetivo de Meta sobre…

    3 Comments
  • Pet SuperAIs

    Or "The Adorable Delusion of Domesticated Superintelligence" If you haven't read Meta's new goal of Personal…

    2 Comments
  • Preparandose para la Artificracia

    Este es la palabra que acuño en castellano como título de mi quinto libro, una mezcla de la palabra del latin…

    1 Comment
  • Getting ready for Artificracy

    This is the term I used as the title of my 5th book is a mix of the Latin word “artificialis” and the Greek word…

  • El Examen E-AGI

    1.- ¿Qué es el Examen E-AGI? 2160 “pruebas” que un robot humanoide debe superar para ser considerado estar a un 100 %…

  • The E-AGI Exam

    1.- What is the E-AGI Exam? 2160 “proofs” that a humanoid robot should “pass” to be considered 100% or beyond “human”…

    1 Comment
  • Haciendo America E-AGI

    La gran "guerra comercial" del siglo comenzó con una falta de preparación sin precedentes por parte de la mayoría de…

    12 Comments

Explore content categories