Throughout the week of March 6–10, Montevideo was decked out to impress because it played host to Khipu 2023, the premier Artificial Intelligence conference within the region. Khipu has a significant mission of promoting progress and development in Latin American AI through student training and constructing a strong AI community. For a whole week, students, researchers, and professionals from across the region gathered to exchange worthwhile experiences and knowledge.

Greater than from different , including from Latin America and around the globe, participated within the event, making this edition of Khipu truly regional and diverse.

This week was jam-packed with inspiring lectures, tutorials, research and highlight talks, covering the newest research topics in AI, including Generative AI, Graph Neural Networks and AI Fairness. Attendees had the privilege of listening to first-level speakers like Nando de Freitas, Samy Bengio, and Peter Norvig, who spoke at length about these and lots of more topics. With their impressive knowledge and expertise, this remarkable speakers lineup provided invaluable insights into the long run of AI.

As if this wasn’t enough, the principal firms within the region that work in AI were there, giving to students immersing experiences and a first-hand sight of what it looks wish to work on the industry developing AI systems.

At we love collaborating with the AI community and we define ourselves by nature, as strong supporters of such honorable initiatives like this. For that reason and as we did in 2019, we supported once more this amazing cause.

After this long introduction, on this post, we’ll cover essentially the most memorable moments of Khipu 2023 and share the fundamental takeaways from this extraordinary conference, through the eyes of our lucky Khipu attendees.

Khipu caught the eye of Computer Vision practitioners on day one with a remarkable recap on this field by Jorge Sánchez: Today’s buildings blocks of visual perception. From the early Hubel and Wiesel experiment (1959) to the current Convolutional Neural Networks models and Multi-Head Attention, explained in easy words.

On day two, Ruben Villegas from Google Brain gave us an prolonged lecture on generative models applied for image and video generation. He began the talk mentioning early approaches like GANs, VAEs, and Normalizing Flows and concluded going deep into two state-of-the-art approaches: Diffusion Probabilistic Models and Vision Transformers. Don’t miss Ruben’s talk Generative Models to go deeper on this fascinating field.Also take a look at the Practicals Sessions for a hands-on experience on this area.

Also, take a look at Parallel Sessions on day 4 and Practicals Sessions for a hands-on experience on this area.

Khipu was no stranger to the generative AI boom of the past few months. Rodrigo Nogueira opened the sphere on day one, with a highlight talk centered across the benefits and downsides of using Large Language Models for information retrieval. Minutes later Joan Bruna Associate Professor at Latest York University, perfectly complemented with a lecture focused on the fundamental challenges behind constructing Foundations Models.

On day two, professor Kyunghyun Cho dazzles on stage together with his outstanding introduction lecture on Natural language processing. In only one hour, he covered the relevance of language in human evolution, the intuitions behind language representation, and straightforward strategies corresponding to the co-occurrence matrix, to Transformers and Attention. He concluded the lecture by declaring that — Statistical patterns cannot distinguish between and . For that reason, researchers must the statistical paradigm for True language understanding — A lot truth…

By the tip of the third day, Nando de Freitas took the stage wearing the Uruguay soccer team jersey and led us on an in depth but very clear journey through Large Language Vision Models. With none difficulty he overflew Transformers and Self-Attention, passing through VQ-VAE to finally end with Diffusion Models. For every of those models architectures, mentioning incredible implementations which have woow us within the last weeks.

On day 4 and as a part of a parallel session, Maria Lomeli who’s a researcher at Meta Research, explained ATLAS: a Retrieval Augmented Language Model (RALM). Unfortunately the video of presentation just isn’t available but you possibly can read more in Atlas: Few-shot learning with retrieval augmented language models.

For those interested in RL, the talk Reinforcement Learning I by Pablo Samuel Castro from Google Brain is the place to go. Pablo provided a comprehensive introduction to Reinforcement Learning, covering fundamentals corresponding to reward and policy functions, value and policy iteration, implementation challenges, and the importance of balancing exploration and exploitation. He also mentioned worthwhile educational resources to delve deeper into Deep RL: an prolonged Introduction to reinforcement learning by himself (Khipu’s presentation is a summary), Spinning Up in Deep RL by OpenAI, and the Dopamine prototyping tool by Google. Don’t miss Pablo’s lecture for more suggestions!

Doina Precup, a professor at McGill University and researcher at DeepMind, continued with Reinforcement Learning II. On this lecture, Doina tackled the issue of the best way to do RL at Scale through Function Approximation and covered some classic implementations. Watch Doina’s presentation to dive deeper.

To see RL in practice, check the at the sensible sessions. Also, within the context of parallel sessions, several researchers complemented the RL experience, sharing their worthwhile knowledge and expertise on the sphere.

One other central topic on the conference was Graph Neural Networks (GNN). On day two, Alejandro Ribeiro presented several large-scale problems (corresponding to authorship attribution and suggestion systems) typically modeled with graphs and solved using Machine Learning techniques applied to Graphs. He also identified that image CNNs, and RNNs for time series, share the identical idea if we expect of them as graph models! In other words, we are able to generalize the convolution concept to a graph and fit a Neural Network (the fundamental idea of GNNs). Don’t miss Alejandro Ribeiro’s lecture.

Later, Gonzalo Mateos continued with Graph Neural Networks II: historical evolution of the sphere and scenarios where GNNs are used for supervised, semi-supervised and unsupervised learning. Gonzalo emphasized that to acquire a satisfactory modeling of the issue, it is important to grasp what the links in our graph mean, by way of our original problem. Gonzalo, citing René Vidal, concluded that we began optimizing features, and now we optimize architectures. He ended the lecture by introducing state-of-the-art GNN solutions.

GNN also had a relevant spot during practical sessions with a comprehensive tutorial on the sphere (including theoretical and practical remarks). Throughout the tutorial, students had the prospect to model a collaborative filtering problem using Graphs after which use a GNN to predict customer reviews over products.

Reflecting on how we work, the objectives and behavior of some ML models, the role of girls in AI, and the role of Latin America in AI had substantial relevance on this edition of Khipu. Specifically, Sara Hooker Director at Cohere, opened the discussion on day three with an enlightening discuss Ethics and Fairness in AI: protected features bias, detecting minority groups to measure fairness, hallucinations, and misinformation of LLMs, and challenges in constructing fair models, corresponding to model drifts or catastrophic forgetting.

Moving on, the sensible session on the Social Impacts of Artificial Intelligence was a possibility to see this topic in motion. Particularly to explore how stereotypes encoded in Transformers models may end up in discriminatory behavior. Later, the Ethics in AI Panel formed by Paola Ricaurte, Sasha Luccioni, Andrés Morales, and Laura Ación, added diverse and worthwhile views to the subject.

Undoubtedly the icing on the cake was a the statement on the impact of Artificial Intelligence in Latin America, signed by many on the organization Committee after which, conference attendants (read the statement here).

Along with keynotes and research talks, Khipu provided an impressive hands-on experience (some already cited on this post). The practical sessions were led by experts in the sphere, covering last trends in AI and providing invaluable insights and best practices for participants and AI practitioners. Here’s a summary of them:

• : Hands-on with JAX, Optax and Haiku: high-performance libraries for computation, optimization, and model learning. Explore these libraries training Convolutional Neural Networks (CNN), from scratch.
• : Explore the Hugging Face hub and unlock the ability of working collaboratively, reusing open data, publishing your datasets as open data, and constructing on top of one of the best open-source models.
• : Learn the fundamentals behind the Transformers architecture and see in greater detail the eye mechanism. Finally, construct all the architecture block by block.
• : Recap Graph Theory after which learn Graph Neural Networks (GNNs): how it really works from a high level, some popular implementations, and the way they work in practice.
• Outstanding code camp, providing invaluable insights and best practices for tackling time series forecasting problems. The tutorial was created by Tryolabs friends in collaboration with Google fellows, making it an exceptional learning opportunity.
• : Explore various RL approaches for solving the classic CartPole, an inverted pendulum system, where an agent must learn to balance a vertical pole by displacing the cart.
• : Walk through the challenges in developing an efficient generative model, specifically for Denoise Diffusion Models (a.k.a. a Rating-Based Generative Model), the backbone of the recent and exciting Dalle-2 and Imagen models.
• : Learn to make use of different tools to evaluate how stereotypes may end up in discriminatory behavior in language technologies and reflect on their social impacts.

Many interesting and enlightening research works were presented during research and highlight talks. Unfortunately, we are able to not mention each presentation on this post, or this might not be a summary. For that reason, we need to make two honorable mentions and encourage you to also review these presentations and parallel sessions.

• : Perhaps unnoticed, because it was considered one of the last Research Talks on day 4, but price every minute of your attention. Samy put together different research problems that Apple is working on each in company products and for the common good of society, with long-term expectations (3–5 years). The fundamental topics are: Self-supervised learning, continuous pseudo-labeling for speech recognition, debiasing large language models, and learning to reason.
• She is a Professor of Computer Science at CICESE (Baja California, México). In a Highlight Talk, she presented two Inspiring stories about AI within the context of her research work. : Children with ASD use a distinct force amount and perform different gesture patterns interacting with haptic interfaces. On this context, Digital Biomarkers and straightforward Machine learning models corresponding to Random Forest may very well be a novel approach to early diagnosis of Autism with high Precision. Need to know more, watch the presentation here.

One other of the high points of the week was the sponsors room, where each company stood out with extremely original stands, and interactive activities to impress students and create a relaxed atmosphere to simply talk and meet latest people. From Machine Learning quizzes, free raffles to an interactive squat wars app that count all of the squats you possibly can do in 30 seconds.

We wish to thank all the scholars, other sponsors and attendees on the whole, who got here to our stand to speak, especially those that said that our stickers were amazing 🙌. ️Finally, congratulations once more to the winners of the free raffles. See you in Khipu 2024 with more raffles and latest stickers designs 💪💪️

— We’re nearing the tip of this post, almost there! —

On Thursday night, Khipu hosted the Women in AI reception, an event intended to unite women in the sphere and foster, encourage and support greater diversity in the sphere of AI within the Latin American community.

With one of the best views of Montevideo as living background, drinks and chill music, the organizers brought together a panel of 4 remarkable women: Paula Martínez, Aiala Rosá, Magdalena Fuentes and Sara Hooker. This extraordinary group of girls inspired attendees, speakers, and special guests with their personal profession journey and experiences, to finalize with a Q&A session. As well as, the organization took the chance to acknowledge the skilled trajectory of Alicia Fernández and Dina Wonsever, two local pioneers in the sphere and emblematic professors from UdelaR.

This evening was definitely, the top of the conference, filling the audience with a renewed sense of empowerment and making a relaxed environment for engaging in stimulating conversations and cultivating latest bonds.

The organizers reserved a really special surprise for the tip. No Te Va Gustar, a well-liked Uruguayan band that sold out stadiums, surprised the audience with an intimate show. The band played several hits and accomplished the show by playing the long-lasting song No Era Cierto — My next memory could also be distorted by drinks and the adrenaline (keep that in mind). I swear I saw my colleague Jairo bravely encourage everyone within the audience to assemble an enormous pogo (mosh pit). Even the conference speakers were there!! — Would Or not it’s crazy if I say then, I saw Jairo starting the nerdiest pogo in history?

(I feel is nice material for a t-shirt no less than…)

The attractive Teatro Solís, an opera house from 1856 (fundamental stage on the country), was chosen to host the closing ceremony. The event was also prolonged to general public (academia, industry and press) who had the privilege to listen to from leading researchers and experts the newest advances and trends in AI.

Three researchers from Latam were in command of raising the curtain. Amongst these presentations, Martín Rocamora stands out definitely, presenting Generative AI in music, candombe drum line on live included! Later Nando de Freitas, Sara Hooker and Kyunghyun Cho dazzled on stage with different presentations about recent advances in AI.

Later, Peter Norvig, a distinguished educator within the Artificial Intelligence field, researcher at Google, and co-writer of the long-lasting book Artificial Intelligence: A Modern Approach (considered one of these books that we must always all read once), took the stage. In his keynote, Peter presented a software engineering journey, from the past’s logic and math-focused languages to the current LLM-based coding assistants like Github-Copilot, ruled by probabilistic and observational strategies. Coding example included!

Concluding with per week full of data and experiences, the ceremony ended with a round table on How and Why We Ought to be Fostering AI in Latin America. The panel of experts in command of guiding the discussion were: Jocelyn Dunstan (Assistant Professor at Universidad Católica de Chile), Fabrizio Scrollini (Executive Director at ILDA), Peter Norvig (Stanford Fellow and Researcher at Google) and Sebastian Barrios (SVP of Technology at Mercado Libre).

Did you miss it? Need to experience it again? Watch the closing ceremony here.

This edition of Khipu was special and incredible in every aspect: from bringing together a number of the brightest minds in the sphere of Artificial Intelligence, bringing students from throughout Latin America, designing per week filled with priceless lectures and tutorials, to the extraordinary effort to have essentially the most amazing closing ceremony and Women in AI reception we could ever have dreamed of.

An standing ovation to the organizing committee, which really surpassed itself remarkably and gave us a firstclass conference.

As we conclude this journey, we are saying goodbye to this edition of Khipu, with joy from the bonds we’ve forged and the friendships we’ve cultivated. Full of inspiration drawn from the fascinating profession paths and private journeys in AI we’ve come across. We’re also very happy with the incredible work that researchers and young talents are doing within the region. Ultimately, we depart with a renewed spirit, committed greater than ever to fostering Artificial Intelligence in Latin American community that just isn’t just strong and united, but resoundingly diverse.