745: 2024 Data Science Trend Predictions

Episode Summary

Executive Summary: Sadie St. Lawrence and John Crohn reviewed 2023 AI/data-science predictions and laid out 2024 trends: rising demand for GPUs and data, LLMs evolving into multimodal assistants/operating systems, slow-thinking reasoning models like Q-Star, enterprise tool consolidation via function-calling, and workforce disruption as AI enables nontechnical users to do analyst-like work. They also discussed regulation, edge AI, and personal productivity.

Main Topics: 2023 prediction recap and validation: They revisited Sadie’s 2023 calls on data as a product, multimodal models, data literacy/accessibility, and AI regulation, noting that all broadly materialized and became more visible in startups, consumer tools, and policy debates. GPU and data-hungry AI infrastructure: Sadie predicts continued demand for AI hardware accelerators and training data, with pressure driving innovation beyond NVIDIA and deeper investment in fine-tuning and niche datasets. LLMs as a new operating system / assistant layer: The discussion frames AI assistants as a platform shift from desktop and mobile, with multimodal systems replacing many screen-and-click interactions through voice, vision, and conversational interfaces. Slow-thinking reasoning models and Q-Star: Sadie argues 2024 will bring breakthroughs in models that reason more deliberately, improving math and logic with less data; John adds context on Q-Star as a possible deep-reinforcement-learning advance. Enterprise consolidation through function calling and tool glue: LLM function-calling and API orchestration are expected to unify fragmented enterprise stacks, reducing the number of point tools and making workflows faster and more integrated. Workforce upheaval and changing data roles: They debate how AI will reshape data science, analysis, and engineering jobs, with more business users doing their own analysis while human roles shift toward higher-value, less siloed work. Regulation, edge AI, and market correction: The conversation also covers global AI regulation, possible AI-bubble overheating, and breakthroughs in on-device/edge AI that could expand use cases in healthcare, agriculture, and retail.

Key Arguments: 2023 predictions largely came true, especially around data as a product, multimodal generation, increased data literacy, and regulatory attention. Demand for GPUs and other accelerators will keep rising because current LLM gains still scale predictably with more compute and more data. Innovation may spread beyond NVIDIA as startups target hardware bottlenecks and energy/latency constraints. LLMs are becoming a conversational interface layer that can see, hear, and eventually replace many traditional UI flows. Reasoning-focused models may be the next major leap because current LLMs still underperform on math and formal logic. Function-calling LLMs will consolidate enterprise stacks by acting as a glue between systems, data, and users. Work roles will become less siloed: analysts, data scientists, and business users will increasingly share AI-assisted analytical capabilities. Regulation is necessary, but overregulation could entrench incumbents and slow startups/open-source communities. Edge AI will matter because running models locally could unlock private, low-latency applications in sectors with limited data connectivity. Personal productivity and self-observation can be improved by systematically tracking behavior, as Sadie does with her Observer method.

Data Points: Podcast episode: 745 - Episode number for this data science trends special. Sadie’s prior trend episodes: 3 consecutive years - John notes this is the third straight annual predictions episode with Sadie. Women in Data community size: 60,000+ women across 55 countries - Sadie’s founder/CEO role and community reach. Sadie’s content audience: over half a million students - Her data science and machine learning instruction reach. Cloud skills requirement in job postings: over 30% - Sponsor copy noting cloud skills demand in ML/data science roles. CloudWolf program length: 30 minutes a day for 30 days - Sponsor offer for AWS certification prep. AI executive order length: 150+ pages - John references the U.S. AI executive order as lengthy and hard to read. LinkedIn post impressions: 14,000 - John’s announcement post about the episode generated significant attention. LinkedIn post reactions: 150 - Audience response to the episode announcement. ChatGPT subscription price: $20/month - John argues the GPT-4 Plus subscription is worth paying for. Generative AI training video length: 2 hours - John references his free YouTube training on LLMs. Half-hour increments: 10 years of data - Sadie’s Observer method tracks her life in 30-minute blocks over a decade. Turing test year: 1950 - John corrects and clarifies the history of the Turing Test. Vision Pro price: $3,500 - John references Apple’s upcoming headset as an example of spatial/AI hardware.

Pivotal Quotes: "I think we’re going to start to see some innovation in the hardware space." — Sadie St. Lawrence: On the first 2024 prediction: rising demand for GPUs and other AI accelerators. "I think this is where we can possibly reach some additional scientific breakthroughs." — Sadie St. Lawrence: On slow-thinking models and why reasoning-focused AI matters for health, physics, and biology. "The way that we work, just as the same way that mobile came out and apps came out, the way we are working today is greatly going to be disrupted by LLMs." — Sadie St. Lawrence: On LLMs becoming a platform shift and operating-system-like layer.

Implications: Expect faster AI adoption, more competition in hardware and models, and major workflow changes. Professionals should build AI literacy now, while organizations should prepare for tool consolidation, regulation, and role redesign.