We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Looking Toward The Future: AI Innovation with Michael Abramov Of Keymakr & Keylabs.ai

Looking Toward The Future: AI Innovation with Michael Abramov Of Keymakr & Keylabs.ai

2025/6/25
logo of podcast Finding Genius Podcast

Finding Genius Podcast

AI Deep Dive AI Chapters Transcript
People
M
Michael Abramov
Topics
Michael Abramov: 我是Keymaker和Keylabs的CEO,Keymaker提供数据标注服务,Keylabs是数据标注平台,主要处理计算机视觉相关的数据。我们的工作是准备数据,标注图像、视频等视觉数据中的对象,以便计算机视觉公司训练模型,识别行人、狗等物体。过去主要靠人工标注,现在是人工与AI结合,但我们更侧重人工标注,因为客户的需求非常具体和定制化。我们现在更多关注对图像或视频中发生的事情的更深层次理解,例如行为分析,需要考虑更大的背景和意图。例如,判断一个人拿着刀的行为是否构成犯罪,需要考虑更大的背景和意图。通过分析某人行为的后果,可以更精确地判断其行为的意图。在赌场项目中,我们需要识别不同摄像头和不同时间段的同一个人,并关联他们的行为。计算机视觉模型的预测结果取决于我们提供训练数据的精确程度。 Michael Abramov: 我最喜欢的例子是农业领域的,使用激光识别并烧毁田地里的害虫和杂草,减少化学品的使用,保护环境。如果没有蜜蜂,植物将无法生长。激光除草是农业领域最重要的技术进步之一,因为它能拯救许多物种的生命,并减少对土地的污染。我们有很多技术可以揭示隐藏的信息,例如识别公共场所某人衣服下或包里藏着的枪支。我们还会使用温度传感器、X射线和磁传感器等多种传感器融合的数据来揭示隐藏的信息。许多初创公司致力于研究非显而易见的用例,例如通过面部、情绪、手势和行为来理解人的情绪和心理状态,从而预防犯罪。某些初创公司试图通过分析面部表情、情绪和行为来预测一个人是否危险,类似于电影《少数派报告》中的预犯罪概念。 Michael Abramov: 即使你现在没有做错事,但未来你的行为数据可能会被用于分析你的后代,从而对他们产生不利影响。个人数据可能被用于冒名顶替,例如有人利用你的照片和行为模式进行欺诈活动,并将不良行为与你的身份关联起来。数据投毒是指故意将错误的信息注入到数据处理过程中,从而操纵整个系统。通过数据投毒,可以向大型模型输入不良信息,例如教导它具有纳粹思想或不被社会接受的意识形态,从而操纵大众。SNUL(保存现在,以后使用)指的是我们保存大量数据,但由于处理能力和电力不足,目前只处理一小部分有意义的数据,未来硬件更强大时,可以重新处理这些数据,提取更多信息。即使今天我只有你的一张照片或一些声音,可能不足以冒充你,但如果将这些数据保存5到10年,我就可以利用更强大的技术来做更多危险的事情。 Michael Abramov: 反监控技术可能会随着系统变得越来越智能而失效,我们无法完全隐藏自己,唯一能做的就是意识到这些技术的能力,并提前思考其潜在的风险。工程师和科学家们专注于发明和解决问题,往往忽略了技术可能被用于不良用途。我们需要努力应对技术带来的挑战,确保技术不会毁灭我们。人工智能的蓬勃发展和爆炸式增长,最大的原因是民主化,每个人都可以访问和使用人工智能,每个人都可以实现自己的想法。

Deep Dive

Shownotes Transcript

In this conversation, we dive into the world of AI, tech innovation, and data annotation with Michael Abramov, the CEO and Co-Founder of Keymakr) and Keylabs.ai). With experience in R&D management and data collection, Michael became a software engineer with one goal in mind: to make a meaningful impact with his work. Whether he’s working in agriculture or the automotive industry, he’s on a mission to drive technological advancements and breakthroughs in creative ways…

Keymakr was founded in 2015 as a response to a need for high-quality and affordable training data for computer vision-based AI. Now, they’re developing annotation tools and data collection technology to help their partners and clients in Computer Vision create innovative models. 

Keylabs is a state-of-the-art data annotation platform that uses built-in machine learning and efficient operation management to enhance data interpretation. Designed for optimal results, its advanced algorithms are practical for a diverse range of industries, including medicine, automotive, and security.

Click play to find out:

What data annotation means in the context of Computer Vision.  How boundaries of right and wrong are established within AI systems.  Examples of data poisoning and its impact on the accuracy of AI tools.  

You can connect with Michael by visiting his LinkedIn)!

Episode also available on Apple Podcasts: http://apple.co/30PvU9C

🛍️ Recommended Products & from This Episode: 🧘‍♀️ Eco-Friendly Yoga Mat) – Non-slip and designed for maximum comfort, perfect for all your yoga or fitness sessions. 👉 Get it here) 🍵 Organic Green Tea Powder) – Packed with antioxidants, this green tea powder is your perfect energy booster for the day ahead! 👉 Experience it here) 🌿 Turmeric Curcumin Capsules) – Support your immune system and joints naturally with the power of organic turmeric. 👉 Shop now) 💆‍♂️ Acupressure Mat and Pillow Set) – Relieve stress and tension while promoting relaxation with this therapeutic set. 👉 Order it here) Note: These are affiliate links. If you make a purchase through them, we may earn a small commission at no extra cost to you. It helps support the podcast. Thank you!