Carmen Amo Alonso: Part 2 (AI Architectures and the Dynamics of Intelligence)

<p>In the second part of my interview with the incredibly talented Carmen Amo Alonso we discuss her fascinating work at the intersection of AI and control, examining topics such as:</p><p>Using tools from control theory to analyse and design AI systems</p><p>Controlling large language models to guarantee a particular sentiment for their output</p><p>Using vision-language-action models to control robots using language</p><p>and a whole lot more!</p><p> </p><p>You can find the transcript for this episode at: <a href="https://docs.google.com/document/d/1ph1yaQuV_ybRM8h9rU5Uvmz9eB5Ar4HVhcBnc7b7zJ8/edit?usp=sharing" target="_blank" rel="ugc noopener noreferrer">https://docs.google.com/document/d/1ph1yaQuV_ybRM8h9rU5Uvmz9eB5Ar4HVhcBnc7b7zJ8/edit?usp=sharing</a></p><p>Follow the podcast on LinkedIn: <a href="https://www.linkedin.com/company/coffee-and-control-podcast/" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/company/coffee-and-control-podcast/</a></p><p>Or connect with me directly: <a href="https://www.linkedin.com/in/lucy-hodgins-733a30175/" target="_blank" rel="ugc noopener noreferrer">https://www.linkedin.com/in/lucy-hodgins-733a30175/</a></p><p> </p><p><strong>References</strong></p><p>The architecture of language: <a href="https://www.youtube.com/watch?v=ApC0AWWTajU" target="_blank" rel="ugc noopener noreferrer">https://www.youtube.com/watch?v=ApC0AWWTajU</a></p><p>Diffusion models: <a href="https://the-principles-of-diffusion-models.github.io/" target="_blank" rel="ugc noopener noreferrer">https://the-principles-of-diffusion-models.github.io/</a></p><p>Skip connection paper: <a href="https://arxiv.org/abs/2410.10609" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2410.10609</a> </p><p>Review of state-space models in AI: <a href="https://arxiv.org/abs/2403.16899" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2403.16899</a> </p><p>3 Blue 1 Brown video on transformers: <a href="https://www.youtube.com/watch?v=wjZofJX0v4M" target="_blank" rel="ugc noopener noreferrer">https://www.youtube.com/watch?v=wjZofJX0v4M</a></p><p>Dynamical systems framework paper: <a href="https://arxiv.org/abs/2405.15731" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2405.15731</a> </p><p>Designing sequence models: <a href="https://arxiv.org/abs/2510.09389v1" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2510.09389v1</a></p><p>Controlling LLM output: <a href="https://arxiv.org/abs/2405.15454v3" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2405.15454v3</a></p><p>Controllability of LLMs: <a href="https://arxiv.org/abs/2601.05637v1" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2601.05637v1</a></p><p> </p><p>Vision-language-action model paper: <a href="https://arxiv.org/abs/2603.05487v1" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2603.05487v1</a></p><p>NARRATE: <a href="https://arxiv.org/abs/2403.10762" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2403.10762</a></p><p>DEMONSTRATE: <a href="https://arxiv.org/abs/2507.12855v1" target="_blank" rel="ugc noopener noreferrer">https://arxiv.org/abs/2507.12855v1</a></p><p>Domitilla Del Vecchio: <a href="https://meche.mit.edu/people/faculty/ddv@MIT.EDU" target="_blank" rel="ugc noopener noreferrer">https://meche.mit.edu/people/faculty/ddv@MIT.EDU</a></p><p>Biomolecular feedback systems book: <a href="https://www.fbswiki.org/wiki/index.php/Biomolecular_Feedback_Systems" target="_blank" rel="ugc noopener noreferrer">https://www.fbswiki.org/wiki/index.php/Biomolecular_Feedback_Systems</a></p><p>Mathias Foo episode: <a href="https://open.spotify.com/episode/1G3vhRoeenTWRrEwauT0ne" target="_blank" rel="ugc noopener noreferrer">https://open.spotify.com/episode/1G3vhRoeenTWRrEwauT0ne</a></p><p>inControl episodes: Mario di Bernardo: <a href="https://www.incontrolpodcast.com/1632769/episodes/19167890-ep44-mario-di-berna