Reverse problem

<p><a class="wikilink" href="/modelling_systems/">Modelling systems</a>, mostly useful in computing properties, is often done as a <strong>forward problem</strong>, which means calculating something from <em>first principles</em>. This is how science has always worked: abstract the problem as much as possible until it becomes useful to make a prediction. </p> <p>On the other hand, we can reverse the question: what is the starting point for a given output? </p> <p>A trivial thing: at what time do I have to leave to arrive on time? </p> <p>As problems get more complex, the reverse problem becomes intractable. This is a crucial aspect of <a class="wikilink" href="/cryptography/">cryptography</a>: it's trivial to encode something, but impossible to decode without the key (<a class="wikilink" href="/physical_unclonable_function/">Physical Unclonable Function</a> are in the same train of thought). </p> <p>Specifically, <a class="wikilink" href="/material_discovery/">material discovery</a> often relies on using <a class="wikilink" href="/dft/">DFT</a>, a powerful (and complete) numerical method to calculate molecular behaviour. Lot's of work has been put into calculating <em>properties</em> based on DFT modelling (which is not at all a trivial step). </p> <p>There are entire databases with modelled systems and interactions ([[<a class="litnote tooltip" href="/literature/@abed2024">@abed2024<span class="tooltiptext">Open Catalyst Experiments 2024 (OCx24): Bridging Experiments and Computational Models</span></a>|Open Catalyst Experiments 2024 (OCx24): Bridging Experiments and Computational Models]], for example). </p> <p>But those still work forward with experimental validation. </p> <p>What if we ask ourselves: "How do I make a material that has X, Y, Z properties, with W boundary conditions?" </p> <p>The problem is intractable today. </p> <p>I believe that <a class="wikilink" href="/tackling_the_reverse_problem_with_quantum_computers/">tackling the reverse problem with quantum computers</a> will be the true breakthrough and in the meantime we need to leverage what we can, including <a class="wikilink" href="/ai_for_the_reverse_problem/">AI for the reverse problem</a>.</p>