900: 95-Year-Old Annie on How to Stay Healthy and Happy

Researcher at Microsoft Robert Usazuwa Ness talks to Jon Krohn about how to achieve causality in AI with correlation-based learning, the right libraries, and handling statistical inference. When dealing with causal AI, Robert notes how important it is to keep aware of variables in the data that may mislead us and force inaccurate assumptions. Not all variables will be useful. It is essential, then, that any assumptions are grounded in a deeper understanding of how the data were gathered, and not what appears in the dataset. Listen to the episode to hear how you can apply causal AI to your projects. Additional materials: ⁠⁠⁠⁠⁠⁠⁠⁠www.superdatascience.com/907⁠⁠⁠⁠ This episode is brought to you by Trainium2, the latest AI chip from AWS and by the Dell AI Factory with NVIDIA. Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

The moral and ethical implications of letting AI take the wheel in business, as revealed by Anthropic: Jon Krohn looks into Anthropic’s latest research on how to use and deploy LLMs safely, specifically in business environments. The team designed scenarios to test the behavior of AI agents when given a goal and a set of obstacles to reach it. Those obstacles included 1) threats to the AI’s continued operation, and 2) conflict between the AI’s goals and the goals of the company. Hear Jon break down the results of this research in this Five-Minute Friday. Additional materials: ⁠⁠⁠⁠⁠⁠⁠www.superdatascience.com/908⁠ Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

“Intelligence has many forms,” says Zohar Bronfman, who speaks with Jon Krohn about the fascinating intersection between computational neuroscience and philosophy, and how it has brought him closer to understanding what is necessary to develop human-like intelligence in machines, as well as his motivations for launching Pecan AI and why predictive models outstrip generative models in business.  Additional materials: ⁠⁠⁠⁠⁠⁠⁠www.superdatascience.com/907⁠⁠⁠ This episode is brought to you⁠⁠⁠ by, ⁠⁠⁠⁠Adverity, the conversational analytics platform⁠⁠⁠⁠ and by the ⁠⁠⁠⁠Dell AI Factory with NVIDIA⁠⁠⁠⁠. Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information. In this episode you will learn: (03:47) Why LLMs aren’t bringing us closer to AGI (33:44) About Pecan AI (51:03) Why data modeling is so challenging (1:01:25) How Pecan AI makes its tools widely accessible

Jason Corso speaks to Jon Krohn in this Five-Minute Friday all about Voxel51’s latest tool, Verified Auto-Labelling, and the company’s incredible success in developing popular tools for computer vision. Additional materials: ⁠⁠⁠⁠⁠⁠⁠www.superdatascience.com/906⁠ Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

RAG LLMs are not safer: Sebastian Gehrmann speaks to Jon Krohn about his latest research into how retrieval-augmented generation (RAG) actually makes LLMs less safe, the three ‘H’s for gauging the effectivity and value of a RAG, and the custom guardrails and procedures we need to use to ensure our RAG is fit-for-purpose and secure. This is a great episode for anyone who wants to know how to work with RAG in the context of LLMs, as you’ll hear how to select the best model for purpose, useful approaches and taxonomies to keep your projects secure, and which models he finds safest when RAG is applied. Additional materials: ⁠⁠⁠⁠⁠⁠www.superdatascience.com/905⁠⁠ This episode is brought to you⁠ by, ⁠⁠⁠Adverity, the conversational analytics platform⁠⁠⁠ and by the ⁠⁠⁠Dell AI Factory with NVIDIA⁠⁠⁠. Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information. In this episode you will learn: (03:28) Findings from the paper “RAG LLMs are Not Safer: A Safety Analysis of Retrieval-Augmented Generation for Large Language Models” (09:35) What attack surfaces are in the context of AI (38:51) Small versus large models with RAG (46:27) How to select an LLM with safety in mind

In this Five-Minute Friday, Jon Krohn reveals how AI is taking on the glitzy world of advertising. Bold claims from Meta and OpenAI contend that users will soon be able to plug in what they want and have AI churn out an ad campaign for little to no cost are shaking the advertising industry to its core. The fact that the four biggest sellers of ads (Google, Meta, Amazon, and ByteDance) are digital companies and accounted for over half of the global market in 2024 adds salt to the wound. Hear the three ways that AI is disrupting the industry, and who (or what) has the most influence on digital consumers to date. Additional materials: ⁠⁠⁠⁠⁠⁠www.superdatascience.com/904 Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

Has AI benchmarking reached its limit, and what do we have to fill this gap? Sinan Ozdemir speaks to Jon Krohn about the lack of transparency in training data and the necessity of human-led quality assurance to detect AI hallucinations, when and why to be skeptical of AI benchmarks, and the future of benchmarking agentic and multimodal models. Additional materials: ⁠⁠⁠⁠⁠www.superdatascience.com/903⁠⁠⁠⁠ This episode is brought to you by Trainium2, the latest AI chip from AWS, by ⁠⁠Adverity, the conversational analytics platform⁠⁠ and by the ⁠⁠Dell AI Factory with NVIDIA⁠⁠. Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information. In this episode you will learn: (16:48) Sinan’s new podcast, Practically Intelligent (21:54) What to know about the limits of AI benchmarking (53:22) Alternatives to AI benchmarks (1:01:23) The difficulties in getting a model to recognize its mistakes

In this episode of “In Case You Missed It”, Jon recaps his June interviews on The SuperDataScience Podcast. Hear from Diane Hare, Avery Smith, Kirill Eremenko, and Shaun Johnson as they talk about the best portfolios for AI practitioners, how to stand out in a saturated candidate market for AI roles, how to tell when an AI startup is going places, and ways to lead AI change in business. Additional materials: ⁠⁠⁠www.superdatascience.com/902 Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

Senior Director of AI Labs for Epiq Lilith Bat-Leah speaks to Jon Krohn about the ways AI have disrupted the legal industry using LLMs and retrieval-augmented generation (RAG), as well as how the data-centric machine learning research movement (DMLR) is systematically improving data quality, and why that is so important.  Additional materials: ⁠⁠⁠⁠⁠www.superdatascience.com/901⁠⁠⁠⁠ This episode is brought to you by the ⁠⁠Dell AI Factory with NVIDIA⁠⁠ and Adverity, the conversational analytics platform⁠⁠⁠. Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information. In this episode you will learn: (05:45) Deciphering legal tech terms (TAR, e-discovery) (13:47) How legal firms use data and AI (29:01) All about data-centric machine learning research (DMLR) (46:58) Lilith’s career journey in the AI industry

“Stay happy and healthy”: In this special Five-Minute Friday, Jon Krohn speaks with Annie, his grandmother, on her 95th birthday. Hear how she is physically and mentally coping with illnesses that limit her mobility and the joys of having a pet. Additional materials: ⁠⁠⁠⁠⁠⁠www.superdatascience.com/900⁠⁠ Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.