https://aibussin.com/tags/forecasting/Recent content in Forecasting on AIBussin — AI applications, systems and booksHugoen-USFri, 10 Jul 2026 11:32:58 +0100https://aibussin.com/post/future/Fri, 10 Jul 2026 11:32:58 +0100https://aibussin.com/post/future/<h2 id="we-trained-a-model-on-editorial-choices-to-see-whether-it-learned-what-happened-next">We Trained a Model on Editorial Choices to See Whether It Learned What Happened Next</h2> <p>Most preference-learning systems use a choice to change the future.</p> <p>A model produces two responses. A human selects one. The chosen response becomes positive evidence, the rejected response becomes negative evidence, and training makes outputs resembling the chosen response more likely.</p> <p>The preference acts as an instruction:</p> <blockquote> <p>Produce more things like this.</p> </blockquote> <p>I wanted to know whether the same choice could also function as evidence.</p>https://aibussin.com/post/preferences/Tue, 07 Jul 2026 00:00:00 +0100https://aibussin.com/post/preferences/<blockquote> <p><strong>A preference is not only a label on what just happened. When the decision belongs to a continuing trajectory, it can also be evidence about what happens next.</strong></p> </blockquote> <h2 id="abstract">Abstract</h2> <p>Most preference-learning systems stop at the choice.</p> <p>A model produces two responses. A human selects one. The chosen response becomes positive evidence, the rejected response becomes negative evidence, and the training system moves on.</p> <p>The work itself usually continues.</p> <p>The selected answer may later be revised, partially retained, contradicted or abandoned. The rejected alternative may reveal a constraint that remains active long after the immediate decision. The preference is therefore not necessarily the outcome. It may be an event inside a longer trajectory.</p>