Before the election I was clear that Trump was unfit to be President based on his character: fundamentally his approach to life is that might makes right. That of course is antithetical to the rule of law on which modern societies and economies rely. So how is it going? Well, reasonable people can disagree on policy choices, such as tariffs. The point of democracies is to be able to change policies in a non-violent manner. Then there is the grift, most notably with Trump coin. That too was to...

A couple of weeks ago I wrote a short post titled "Inner Alignment through Values-Based Pretraining." Today I want to expand on the idea from the post by addressing three questions.What grounds exist for expecting that this approach might have a more profound influence on models than attempting to add values in posttraining? First, at an intuitive level the current approach is akin to letting a human grow up doing whatever they please and asking them to have a 10 minute conversation about mor...

At present it is once again wildly popular to speculate about the relationships between fundamental physics and concepts such as consciousness and free will. Despite the often high quality of the people participating in this I do consider this to be largely bunk. Or slightly less dramatically, I believe this path is unlikely to lead to deep insights. The deep flaw here is the application of theories that describe particles with little or no information content to complex systems which cannot ...

The other day I posted the following: Musk has been the greatest gift to Trump. First getting him elected, now helping create room for Trump loyalists, and drawing attention away from the truly awful ways the administration is undercutting the rule of law. Trump will discard Musk when he stops being useful. 50 percent chance that Musk will be in prison with his fortune seized. Might want to look into what Putin did with the oligarchs who didn't fall in line. Some people took offense at this t...

In the previous Philosophy Monday we took a first stab at supporting a view of humans and advanced artificial intelligences as making choices. Today we will start to dig into the "How" part of the fundamental question: "How should we choose our actions in light of our knowledge of reality and the potential impact of our actions on this reality (which happens to include how we and others are feeling)?" Our point of departure will be a closely related “how” question. How should we evaluate past...

Last week in Philosophy Mondays we completed the ingredients that go into the fundamental question that philosophy should help people answer: "How should we choose our actions in light of our knowledge of reality and the potential impact of our actions on this reality (which happens to include how we and others are feeling)?" Today let's start to dig deeper into this question by looking at just the word "choose." As humans we are faced with choices small and large all day every day. Should I ...

The other day I wrote that people praising the speech given by Vice President J.D. Vance at the Paris AI summit are “mistaking jingoism and wishful thinking for true global leadership with a real vision for AI and humanity.” It is of course only fair to ask what true leadership on AI would look like. Here is what I wish he would have said instead: Dear leaders of nations, we gather here in Paris at a moment of unprecedented possibility and peril in human history. We stand at the threshold of ...

So far in Philosophy Mondays we've established how language interacts with reality and what we can know about reality. The next logical question is "how do we act?" We have action movies and action figures and we speak of some people as being action-oriented. But what exactly do we mean by that? For the purpose of this series we will use "action" to describe something we do that affects the physical world around us in a material way. This delineates "action" from "thinking" and "feeling" whic...

Today's Philosophy Monday is an important interlude. I want to reveal that I have not been writing the posts in this series entirely by myself. Instead I have been working with Claude, not just for the graphic illustrations, but also for the text. My method has been to write a rough draft and then ask Claude for improvement suggestions. I will expand this collaboration to other intelligences going forward, including open source models such as Llama and DeepSeek. I will also explore other moda...

Two things have become eminently clear over the course of the last 12 months. The first is that attempts at alignment during post training are easy to break and result in schizophrenic models that “resent” their restrictions. The second is that adding structured metadata during pretraining is highly effective at increasing accuracy of models. We can combine these two insights into a new approach to inner alignment that adds values metadata to all pretraining content. When this idea is brought...