Deepesh Deepak

Riding the wave of AI, ML, DL, CV, RL, Drones & Robotics. Let's automate the repetitive tasks so that the humanity can march towards higher form of intelligence

[Paul Graham @paulg]

Wow, as recently as 2023 the Duke University library system dropped Basecamp because the founders weren't sufficiently PC. Or as they put it, because of "harms perpetuated" by them. (There have to be harms, or you're just banning speech.)

https://t.co/U2BhrRN1O0

[Boyan Slat @BoyanSlat]

The Ocean Cleanup is looking forward to deploy in India!

Data collection in the rivers of Mumbai is starting in the next few weeks, with the aim to have Plan Mumbai ready for implementation (contingent on funding) by the end of the year.

[Deepesh Deepak @deepeshdeepak2]

@elonmusk @lindayaX Could you please add search functionality in Bookmarks please? Write now we have to scroll way down to see previous bookmarks.

[Deepesh Deepak @deepeshdeepak2]

@elonmusk @lindayaX Could you please add search functionality in Bookmarks. Write now we have to scroll way down to see old bookmarks which is inefficient. Thanks.

[Bojan Tunguz @tunguz]

I wonder what does it feel like to be a normie and work in a field where you don’t have to upskill every couple of hours.

[Alvin Foo @alvinfoo]

The 12 stages in burnout process.

[DeepLearning.AI @DeepLearningAI]

Palantir, an analytics company with military and intelligence clientele, showcased a chatbot capable of analyzing battlefield activity, releasing drones, and suggesting plans of action.

Learn more in #TheBatch: https://t.co/s5Wgf5FABT

[Programmer Humor @PR0GRAMMERHUM0R]

I may need that again someday! https://t.co/hHToRz93ai

[Yann LeCun @ylecun]

IMAGEBIND: One Embedding Space To Bind Them All.

Learns a joint embedding across six different modalities - images, text, audio, depth, thermal, and IMU data.

An open source project by Meta-FAIR.

Paper: https://t.co/g07iDT5rVO
Demo: https://t.co/80OKqOyGEb
Code: https://t.co/J4jrRm3Kik
Blog post: https://t.co/Meh7QSQsRS

[Elon Musk @elonmusk]

WhatsApp cannot be trusted

[Yann LeCun @ylecun]

If some ill-intentioned person can produce an evil AGI, then large groups of well-intentioned, well-funded, and well-organized people can produce AI systems that are specialized in taking down evil AGIs.
Call it the AGI police.

No need to bomb data centers.

[Yann LeCun @ylecun]

In many animal species, evolution encodes objectives.
It is up to the individuals to figure out the behaviors that optimize these objectives.

Encoding objectives rather than behaviors is much simpler, more efficient, and more adaptive way for evolution to specify complex behaviors.

the same way it is much easier for human engineers to specify a loss function and rely on optimization for a system to perform perception or control than it is to design a perception or control system from scratch.

[Santiago @svpino]

TensorFlow or PyTorch?

For educational purposes, I always use PyTorch. Building things from scratch is much more verbose in TensorFlow.

TensorFlow, however, is more mature for production systems. Their TFX framework is great, so that's the one I use professionally.

[Deepesh Deepak @deepeshdeepak2]

Anyone looking for a quick intro to V4L, FFMPEG, and Gstreamer like multimedia utilities, check out:
V4L: A Complete Practical Tutorial https://t.co/jMEcMhjDdB

[Deepesh Deepak @deepeshdeepak2]

I just published CNN Visualization Techniques: Feature Maps, Gradient Ascent https://t.co/nv4c0NcS03

[Deepesh Deepak @deepeshdeepak2]

Another tutorial on the GAN series:
I just published DCGAN implementation from scratch on the FMNIST dataset in PyTorch https://t.co/ziN60yEKuP

[Deepesh Deepak @deepeshdeepak2]

I just published GAN implementation from scratch on MNIST dataset in PyTorch https://t.co/9pjp3WiHIp

[Deepesh Deepak @deepeshdeepak2]

Unloading-the-Cognitive-Overload-in-Machine-Learning:
I just published LeNet-5 Implementation on MNIST in PyTorch https://t.co/d0cqn8qv92