We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode How Sama is Improving ML Models to Make AVs Safer // Duncan Curtis // #307

How Sama is Improving ML Models to Make AVs Safer // Duncan Curtis // #307

2025/4/18
logo of podcast MLOps.community

MLOps.community

AI Deep Dive AI Chapters Transcript
People
D
Duncan Curtis
Topics
Duncan Curtis: 我专注于为全球大型企业赋能AI，无论是在自动驾驶还是生成式AI领域。我们致力于改进机器学习模型的准确性、速度和成本效益，降低模型故障风险，并降低汽车制造商的总拥有成本。我们拥有全职员工和平台，不断思考如何预测未来趋势，并充分利用人工参与。在自动驾驶领域，我们处理来自激光雷达、摄像头、雷达和超声波传感器的大量数据，需要人工标注和解读这些数据，这需要高度的智能和对场景的理解。我们利用AI辅助标注，提高效率，并专注于捕捉对AI模型最重要的信息。我们还关注数据选择和数据价值最大化，通过数据筛选和优化，提高数据利用率。我们也关注数据偏差问题，通过数据收集和标注过程中的多样性和代表性，减少偏差并提高模型鲁棒性。我们处理各种异常情况，例如罕见的交通场景，并及时更新模型以适应新的情况。我们也关注不同地区和国家的驾驶习惯和交通法规差异，并将其纳入模型训练中。在技术发展迅速的背景下，我们也关注技术的前向兼容性，避免因技术更新而导致前期投入的资源浪费。生成式AI的兴起也为我们的业务带来了新的机遇，我们发现我们的现有英语为主的劳动力和训练流程与市场需求非常契合。对于需要专业知识才能标注的数据，我们关注如何利用现有技术和资源来解决问题，并预测未来发展趋势，我们相信未来AI数据标注的需求将转向对更通用的智能和更复杂的AI系统（例如具身AI）的反馈和评估。现有的AI技术已经能够降低AI产品开发的门槛，但产品的差异化将主要体现在产品设计和用户体验上。成功的AI产品不仅仅依赖于底层技术，更依赖于优秀的产品设计和用户体验。数据是当前AI发展的瓶颈，其原因在于数据量巨大、数据质量参差不齐以及数据处理的复杂性。Sama公司为客户提供全面的AI数据服务，从咨询到数据标注和模型训练，帮助客户解决各种AI相关的业务问题。我们参与过利用AI技术进行大象保护的项目，通过识别大象臀部特征来追踪大象。在AI模型验证阶段，需要找到一种平衡的方法，在保证验证质量的同时，降低成本。评估AI项目的ROI至关重要，可以通过MVP和POC等方式快速验证AI项目的价值，并根据业务目标选择合适的评估指标。 Demetrios: 作为主持人，我主要负责引导访谈，并就数据标注、AI模型改进、自动驾驶安全等话题与Duncan Curtis进行讨论。我关注数据作为AI发展瓶颈的问题，以及如何评估AI项目的ROI。我与Duncan Curtis探讨了数据标注中人工参与的重要性，以及如何平衡人工和自动化，以提高效率和降低成本。我们还讨论了AI模型的鲁棒性和前向兼容性，以及如何应对技术快速发展带来的挑战。此外，我还与Duncan Curtis探讨了AI项目价值评估的重要性，以及如何将AI项目与业务目标相结合。

Deep Dive

Shownotes Transcript

How Sama is Improving ML Models to Make AVs Safer // MLOps Podcast #307 with Duncan Curtis, SVP of Product and Technology at Sama.

Join the Community: https://go.mlops.community/YTJoinIn Get the newsletter: https://go.mlops.community/YTNewsletter

// Abstract

Between Uber’s partnership with NVIDIA and speculation around the U.S.'s President Donald Trump enacting policies that allow fully autonomous vehicles, it’s more important than ever to ensure the accuracy of machine learning models. Yet, the public’s confidence in AVs is shaky due to scary accidents caused by gaps in the tech that Sama is looking to fill.As one of the industry’s top leaders, Duncan Curtis, SVP of Product and Technology at Sama, would be delighted to share how we can improve the accuracy, speed, and cost-efficiency of ML algorithms for ​A​Vs. Sama’s machine learning technologies minimize the risk of model failure and lower the total cost of ownership for car manufacturers including Ford, BMW, and GM, as well as four of the five top OEMs and their Tier 1 suppliers. This is especially timely as Tesla is under investigation for crashes due to its Smart Summon feature and Waymo recently had a passenger trapped in one of its driverless taxis.

// Bio

Duncan Curtis is the SVP of Product at Sama, a leader in de-risking ML models, delivering best-in-class data annotation solutions with our enterprise-strength, experience & expertise, and ethical AI approach. To this leadership role, he brings 4 years of Autonomous Vehicle experience as the Head of Product at Zoox (now part of Amazon) and VP of Product at Aptiv, and 4 years of AI experience as a product manager at Google where he delighted the +1B daily active users of the Play Store and Play Games.

// Related Links

Website: https://www.sama.com/Tesla is under investigation: https://www.cnn.com/2025/01/07/business/nhtsa-tesla-smart-summon-probe/index.htmlWaymo recently had a passenger trapped: https://www.cbsnews.com/losangeles/news/la-man-nearly-misses-flight-as-self-driving-waymo-taxi-drives-around-parking-lot-in-circles/https://coruzant.com/profiles/duncan-curtis/https://builtin.com/articles/remove-bias-from-machine-learning-algorithms