Kasparas Aleknavičius on Deep learning and AI developmentHealthcare futurist. Meaning Doctor, AI Developer, Business Developer 🧠🚀Some time ago
Allegory of two distinct Deep Learning approaches in autonomous driving. - Tesla neural networks replicate our brain on a cell level. Decisions are based on tiny predefined details and using all of them. That's what developing brain does. - Comma.ai does the complete opposite - replicating our brain brain function on a holistic interconnected level. Decisions are based on the big picture, using different detail-based arguments and picking the most suitable ones.
What Tesla has in common with baby’s brain? And why it’s a good example of health economics? | by Kasparas Aleknavicius | Dec, 2020 | Medium | The Shadow
medium.com

Appreciate
Comment
Book

Danielius VisockasR&D @ Neurotechnology; Burger geek; Speech & neural nets
One small note. TSLA actually doesn't feed all the different things into their nets. They have a backbone resnet-ish network and multiple heads for different tasks (because of multiple heads they call it the hydra). So the main difference between tesla and comma is single-task vs multi-task learning. More on tsla nets: https://www.youtube.com/watch?v=oBklltKXtDE P.S. Tesla neural nets do not replicate our brain on a cell level. All neural nets are merely loosely inspired by our brains...
4 months ago
·
Appreciate
Kasparas AleknavičiusHealthcare futurist. Meaning Doctor, AI Developer, Business Developer 🧠🚀
Definitely yes! Single task vs multi task. I did not go into the technical details of nets and mentioned “replication” quite carelessly, because my goal was to emphasize the allegory of these approaches. Multi-task learning can be the example of how we start absorbing and processing information when we are born. When growing old our brain starts so process information differently - it becomes a single-task neural net. And I really mean not the exact mechanisms, but the principles in decision making. Short example of what I meant: On a daily basis adult brain becomes unaware of grass beeing green in the summer because it is used to notice it so many times. Our brain has more important things to focus on, depending on the situation.(LSD turns off this mechanism and we start noticing previously unimportant (or filtered by our brain) stimuli). But the baby’s brain notices grass being green in the same manner that it notices everything! And it gives everything an equal amount of importance. (Thats also one of the reasons we dont remember that part of our life). Nevertheless I agree I could have expressed this allegory more clearly, I did not mean to explain technical details, just wanted to give enough to explain the similarities. And you are right about neural nets beeing meerly inspired by our brain. This allegory are just thoughts that came into my mind, but it’s on a more physophycal level since I have never heard anyone saying these kind of things. By no means I am saying that it’s a mechanistic replication, but I see very clear overlaping principles between A) these two approaches and B) two very distinct stages of our brain - infant and adult.
4 months ago
·
Appreciate

Download Qoorio to talk & learn from other Humans
MORE INSIGHTS YOU MIGHT LIKE
Learn more by discovering other insights
Agnes Li on Deep WorkEdtech startup co-founderSome time ago
Tips to avoid distractions of instant messaging (Except app blockers and putting away your phone): 1. Use email instead of instant messaging as much as possible 2. Check your messaging apps once every 24 hours 3. Have weekly / monthly breaks from messaging. Set your WhatsApp status to "not responding to messages today" or set an automatic reply to every message saying that you are away and call if it's urgent.
4

Appreciate
Comment
Book
Agne Nainyte on LEAN Problem Solving Rock StarDigital Transformation | Process Excellence | Women EmpowermentSome time ago
There is an interesting story to listen on Planet Money Podcast about elephants, deep learning and great problem solving approach. We all know that we have a massive elephants poaching problem in Africa. However, we don't know the exact details: how many, where, when, etc. According to LEAN principles, you always start with clarifying and quantifying your problem statement. This is what scientists started to do by listening to the sounds of elephants. However, very soon they had a roadblock - too much data to process. So, they applied deep learning techniques working together with machines and teaching them on how to differentiate sounds. What a great example of how machine working together with humans can help in solving problems. What are your favorite examples how machines collaborate together with humans?
5

Appreciate
Comment
Book
Vaida Saulyte on Business Development LearningsTech | Education | Co-CreationSome time ago
I used to have a narrow view when forming partnerships. The process would like: 1. I need new partners for x project 2. Contact organizations to create y partnership for x project 3. X project over, y partnership over There are many reasons why this does not work out: o It misses out potential opportunities o Does not take a consideration of the other partner's needs and future goals o Poor response rate because of "I never cared about you before I needed something out of you" approach o Focus on quantity instead of quality If you really want to develop a partnership and see potential collaboration opportunities, do not reach out to a partner only when you NEED them. Approach them gently, and with a lot of care. Provide value and create a "social capital" - so that they feel obliged to support you once you need help as you did the same for them in the past. Do not start a partnership from "what can you do for me?" Instead, focus on "what can I do for you?" It will take time, it is a long-term game and the benefits would often not be seen immediately. However, it is the way to build sustainable, valuable and impactful partnerships.
3

Appreciate
Comment
Book
Download Qoorio to talk & learn from other Humans
Sign inJoin Qoorio
We use cookies to personalise content, provide social media features, and analyse our traffic. We value your privacy and only use the most necessary and analytical cookies. You can opt out at any time.