Having been an avid podcast listener the last few years and always an eager reader, I now have a feeling that topics, debates, and conclusions are being recycled over and over again. Also the mix of podcast topics (science, politics, world affairs, philosophy) and selection of podcast guests have started to converge beween the podcasts.
AI
Can I trust an LLM?
LLMs use gigawatts of precise, deterministic computing power to produce human Type 1 thinking while we still need to rely on our noisy, chemical brains for serious Type 2 tasks. The singularity sometimes looks more like a mud pit.
Proof that LLMs hallucinate
I asked ChatGPT if it was prone to hallucinations. It admitted it was.
Wrong and right about AI risks
Anthropomorphizing AI applications is not productive. AI applications are systems and should be designed using systems engineering.
It takes a human to find the dragons
Without intrinsic motivation, the ability to hypothesize beyond training, or the capacity for real-world experimentation, AI will remain a tool that processes human-created information rather than a creator of new knowledge.
Does an AI create new knowledge?
The LLM can interpolate within the knowledge space it’s been trained on, filling in gaps by blending concepts in novel ways. However, it will stay within the convex hull of its training data, constrained by the boundaries of what it has learned.
Open and closed loop learning
Closed-loop learning is almost a law of nature. It is applied by all sustainable natural systems and in most successful human endeavors.
Despite the obvious benefit or even need of closed-loop learning, we fail to implement it in many contexts that would clearly benefit from it.
Subjective experiences from an information differential
A hypothesis that I’d like to investigate is that subjective experiences arise as a consequence of state changes in an algorithm. I provisionally call such state changes information differentials.
Can an AI take responsibility?
A mantra repeated several times at a healthcare conference that I attended recently, is that only humans, not AI, can take responsibility for something. This made me think more deeply about what it really means to take responsibility and what, if anything, sets humans and AIs apart in this respect.
Emergent properties misunderstood
Most systems have the properties they have because they were designed that way, either by humans or by nature, not because they “emerged”.