Articles
To build an AI Conversational Assistant with human-quality capabilities they must have been memorised with appropriate Cognitive Memories.
How do we study ‘objects’ that are effectively unreachable? How can we study the human mind and apply the learning to AI thinking?
Testing Artificial Intelligence and how Deus ex machina is often unwittingly embedded into many test scenarios leading to meaningless results.
Mimicry is not intelligence, this article discusses why.
This paper discusses the idea that a definitive counterexample unambiguously disproves a conjecture and why although Generative AI is very good for somethings it will never achieve AGI.
To the man with a hammer every problem is a nail. For many AI researchers their hammer is Generative AI and to these individuals all cognitive capabilities have the properties of a nail!
The really useful AI feature of Infinite Memory now seems to have been recognised by Microsoft as Mustafa Suleyman (CEO, Microsoft AI) remarked in a recent interview with The Times …
Simple reasoning in a Large Language Model could be either two-hop or shortcut reasoning. Surely a shortcut is a good thing, isn’t it? Here's why it isn't.
The current test sets used to measure AI intelligence have no scientific value and do nothing to show intelligence, we look at why.
As this article highlights the dismissive attitude towards maths can have devastating effects which begs the question, what does the future look like if AI can’t perform simple maths without making errors?
How we could build smarter, lighter, more human-like AI systems that don't require server farms to generate a simple exchange.
The human brain with its low power and instant update capability must use language in a simple way. Here's why breaking down language into a geometric structure allows us to create accurate and verifiable responses from AI.
Are hallucinations a price worth paying for an AI future - even though they create false memories? If we're going to use AI for medicine the answer should be NO.
Science, and the mathematics that underpins it, is the foundation for so much progress in the modern world yet it seems to be ignored when it comes to AI.
There is an increasing AI focus on Agents with long-term memory, and also in AI healthcare applications. However, for safe and successful applications in healthcare precise memories will be essential. …
There continues to be talk of the imminent arrival of AGI / Super Intelligence and in this article we look at why scaling from the achitectures currently used won't work.
A simple introduction to Context and its importance to human-quality conversations in AI.
Are we really close to achieving AGI or is it a case of the Emperor's New Clothes? In this article we explore 'Where is the path to AGI?'