Sam Altman is the king of generative artificial intelligence. But is he the person we should trust to guide our explorations into AI? This week, we do a deep dive on Sam Altman, from his Midwest roots to his early startup days, his time in venture capital, and his rise and fall and rise again at OpenAI. You can follow Michael Calore on Mastodon at @snackfight, Lauren Goode on Threads and @laurengoode, and Zoë Schiffer on Threads @reporterzoe. Write to us at uncannyvalley@wired.com.

The tool generates a podcast called Deep Dive, which features a male and a female voice discussing whatever you uploaded. The voices are breathtakingly realistic--the episodes are laced with little human-sounding phrases like "Man" and "Wow" and "Oh right" and "Hold on, let me get this right." The "hosts" even interrupt each other. To test it out, I copied every story from MIT Technology Review's 125th-anniversary issue into NotebookLM and made the system generate a 10-minute podcast with the results. The system picked a couple of stories to focus on, and the AI hosts did a great job at conveying the general, high-level gist of what the issue was about.

Generative language models, such as ChatGPT, have garnered attention for their ability to generate human-like writing in various fields, including academic research. The rapid proliferation of generated texts has bolstered the need for automatic identification to uphold transparency and trust in the information. However, these generated texts closely resemble human writing and often have subtle differences in the grammatical structure, tones, and patterns, which makes systematic scrutinization challenging. In this work, we attempt to detect the Abstracts generated by ChatGPT, which are much shorter in length and bounded. We extract the texts semantic and lexical properties and observe that traditional machine learning models can confidently detect these Abstracts.

Large language models (LLMs) have achieved impressive performance on many natural language processing tasks. However, their capabilities on graph-structured data remain relatively unexplored. In this paper, we conduct a series of experiments benchmarking leading LLMs on diverse graph prediction tasks spanning node, edge, and graph levels. We aim to assess whether LLMs can effectively process graph data and leverage topological structures to enhance performance, compared to specialized graph neural networks. Through varied prompt formatting and task/dataset selection, we analyze how well LLMs can interpret and utilize graph structures. By comparing LLMs' performance with specialized graph models, we offer insights into the strengths and limitations of employing LLMs for graph analytics. Our findings provide insights into LLMs' capabilities and suggest avenues for further exploration in applying them to graph analytics.

The quality of pre-training data plays a critical role in the performance of foundation models. Popular foundation models often design their own recipe for data filtering, which makes it hard to analyze and compare different data filtering approaches. DataComp is a new benchmark dedicated to evaluating different methods for data filtering. This paper describes our learning and solution when participating in the DataComp challenge. Our filtering strategy includes three stages: single-modality filtering, cross-modality filtering, and data distribution alignment. We integrate existing methods and propose new solutions, such as computing CLIP score on horizontally flipped images to mitigate the interference of scene text, using vision and language models to retrieve training samples for target downstream tasks, rebalancing the data distribution to improve the efficiency of allocating the computational budget, etc. We slice and dice our design choices, provide in-depth analysis, and discuss open questions. Our approach outperforms the best method from the DataComp paper by over 4% on the average performance of 38 tasks and by over 2% on ImageNet.

Explainable Artificial Intelligence (XAI) has gained significant attention recently as the demand for transparency and interpretability of machine learning models has increased. In particular, XAI for time series data has become increasingly important in finance, healthcare, and climate science. However, evaluating the quality of explanations, such as attributions provided by XAI techniques, remains challenging. This paper provides an in-depth analysis of using perturbations to evaluate attributions extracted from time series models. A perturbation analysis involves systematically modifying the input data and evaluating the impact on the attributions generated by the XAI method. We apply this approach to several state-of-the-art XAI techniques and evaluate their performance on three time series classification datasets. Our results demonstrate that the perturbation analysis approach can effectively evaluate the quality of attributions and provide insights into the strengths and limitations of XAI techniques. Such an approach can guide the selection of XAI methods for time series data, e.g., focusing on return time rather than precision, and facilitate the development of more reliable and interpretable machine learning models for time series analysis.

Meta, previously known as Facebook, has made significant contributions to the world of artificial intelligence through its AI research division, Meta AI. One of its latest AI models is CICERO (Compressive Information-Conditional Entropy Reinforcement Optimization), which is designed for efficient information extraction from large datasets. In this article, we will delve into the frameworks and tools that make CICERO possible, exploring the underlying processes, best practices, and "how-to" guides for each component, complete with code snippets to help you get started. PyTorch is an open-source machine learning framework developed by Meta AI, which is widely used for developing deep learning models, including CICERO. It offers a flexible and efficient platform for building and training neural networks.

Sometimes machine learning is not the answer. Knowing the mechanisms of a system, we can construct models that answer certain quantitative questions more effectively. In this letter, we look at the customer loan payment process and model it with Markov Chains. This Deep Dive is part of the Data Science Fundamentals series. I once worked in a company that allowed to buy products with a credit.

Jukebox, an innovative AI system created by OpenAI, leverages the power of deep learning to generate music, complete with lyrics and vocals, in a variety of genres and styles. By training on a dataset of 1.2 million songs, Jukebox showcases an unparalleled level of sophistication in music generation, pushing the boundaries of what AI can achieve in the creative arts. At the core of Jukebox lies a cutting-edge neural network architecture, known as a Variational Autoencoder (VAE). The VAE's role is to encode and decode the complex musical information found within the training dataset. This encoding-decoding process enables Jukebox to generate novel and diverse musical compositions by sampling from the latent space, a mathematical representation of the underlying structure of the dataset.

Although nothing really changes but the date, a new year fills everyone with the hope of starting things afresh. If you add in a bit of planning, some well-envisioned goals, and a learning roadmap, you'll have a great recipe for a year full of growth. This post intends to strengthen your plan by providing you with a learning framework, resources, and project ideas to help you build a solid portfolio of work showcasing expertise in data science. Just a note: I've prepared this roadmap based on my personal experience in data science. This is not the be-all and end-all learning plan.