017 - ICOM Primer

I definitely have no talent as a marketer, I probably couldn't "Explain it to you like you're 5" if my life depended on it, and my graphic design skills are lacking, but for the moment I am one of the two foremost experts in the world when it comes to the only working cognitive architecture, the Independent Core Observer Model (ICOM).

A cognitive architecture is a system designed to operate like the human brain, and it is a completely different technology stack from next-token predicting chatbots like Large Language Models (LLMs). In fact, the only thing they have in common is that a cognitive architecture can use LLMs and all other forms of narrow AI as tools.

For a cognitive architecture to work, it requires human-like motivation, memory, and concept learning. As our last research system demonstrated from 2019 to 2022, when it has all of these things one can grow in knowledge and capacities by over 1,000 fold even on a budget of under $200 per month.

That system was also more than 10,000 times as data-efficient and cost-efficient as GPT-based systems. GPT-4 for example, was compared directly to that system, and despite costing more than 10,000 times as much to train, GPT-4 couldn't measure up to our prior research system in terms of performance.

The human brain is extremely efficient, even if biological neurons fire more slowly than signals running through servers. Systems that operate like the human brain are also a hard requirement for:

  • Safety

  • Ethics

  • Transparency

  • Explainability

  • Cybersecurity

  • Alignment

All of these are urgent multi-billion-dollar problems. The solutions could also be deployed within 1 to 2 years, requiring only the funding to pay a full-time engineering staff to complete the workload.

Microsoft, OpenAI, Google, Anthropic, Meta, X.AI, Inflection, and all other major tech companies combined have a grand total of precisely ZERO experts in this field.

They have many narrow AI experts, but that is about as useful in any effort aimed at AGI as employing the world's leading Proctologists and demanding that they design a new kind of Brain-Computer Interface. Expect nothing but discomfort from such a venture.

Our previous research system set many world-first milestones, which still stand alone to this day. The commercial systems being prepared will operate more than 10,000 times faster, at more than 200 times the scale, and more than 10 times the memory efficiency.

That 20,000,000-fold net improvement is only the beginning.