PodcastLanguage, Graphs, and AI in IndustryClaimPaco Nathan says AI applications in regulated or industrial settings need software engineering, operations, security, legal review and domain expertise rather than only an API call.1:00:42
Threads / Humans in the loop / 2024
Humans in the loop in 2024
16 sessions
PodcastLanguage, Graphs, and AI in Industry
From Research to Production: Fine-Tuning & Aligning LLMsClaimA good instruction dataset needs clear instructions, diverse tasks and topics, consistent formatting, and high-quality human feedback.12:57
Graphs and LanguageClaimGraph visualization can help humans understand black-box models and support expert semi-supervision.1:54
Data Labeling Best PracticesClaimTextMine built its own data labeling team and fine-tuned its own models, and this talk presents its lessons rather than prescribing one correct labeling method.1:07
PodcastReliable LLM Products, Fueled by FeedbackClaimThe drone project used more than 10K high-resolution images that had to be manually annotated before training a custom YOLO 3 model.4:25
Balancing Speed and SafetyClaimSuccessfully deployed large language model applications are often constrained by human review or limited to internal question answering.20:28
Vision and Strategies for Attracting & Driving AI Talents in High GrowthClaimOlga Beregovaya says human reviewers remain necessary for tasks such as assessment, ranking, validation, post-editing, and fact checking because AI models can hallucinate and data can be biased.15:27
The Next Revolution in AI: LLMs and BeyondClaimFor qualitative applications, human review and quick sanity checks are useful for improving prompts and content.8:46
Turn Data Chaos into AI Strategy with Programmatic AI Data DevelopmentPushed backProgrammatic labeling functions should not be used directly as the final inference rules because a rule-based system is not robust or scalable enough for new data.22:35
PodcastHow Agentic Workflows Will Change EverythingClaimFor an initial MVP, developers should keep a human in the loop, inspect the agent's planned actions, and constrain its environment.17:48
Cleric AI SRE: Towards Self-healing Autonomous SoftwareClaimProduction infrastructure is complex, dynamic, and difficult for humans to keep in mind as systems and connections grow.2:57
The Future of Healthcare: AI is HerePushed backShaun disputes the idea that AI should merely match human performance, saying HeyRevia's results show its agents can outperform humans in comparable healthcare phone-call scenarios.23:14
Why Planning is the New SearchPushed backFabian disputes the assumption that agentic workflows should immediately be fully self-driving, arguing that customers generally want human oversight and gradual automation first.26:11
Building Reliable AgentsClaimHuman input should remain part of agentic-system design because these systems are unlikely to complete large tasks reliably every time.19:36