Geeky Insights for AI Enthusiasts
Posts
🤔 Do they need our permission to build AGI?

🤔 Do they need our permission to build AGI?

and will AI clones be joining your meetings?

October 18, 2024

“No man is good enough to govern another man without the other's consent.” - Abraham Lincoln

In today’s newsletter:

Did we consent to AGI?
Finding your by your walk
Security issues worsen with AI
AI Clones in Meetings

Pense

Artificial General Intelligence( AGI) is the question of the week. Sam Altman called it the “magic intelligence in the sky”…creepy right.

But it never occurred to me to question whether he had the right to start building it.

so these week, I encourage you to read Sigal Samuel of vox.com’s thoughts on it.

His thesis: The public did not consent to artificial general intelligence. He responds to objections raised to this under three headings:

Our use is our consent.
The public is too ignorant to tell innovators how to innovate.
It’s impossible to curtail innovation anyway.

Read it here.

Article

In the future, I betting that cops will try to catch criminals by how they walk, or gait recognition. Security systems that have technology that attempts to recognize your face , and recognizing you from your walking is coming. The Chinese government deployed this tech since 2018. Though I’m not sure how it’s going. Read about it here.

Gait recognition is being used as a biometric security measure, but there’s always two sides to everything.

And since I have a mild fascination with cool creepy tech, we’re going to dive into the developments in this space.

Apple is working on a sensor that recognizes your gait using the built-in headphone accelerometer — whatever that is. Read more here. Someone on the team is a Michael Jackson fan. It’s named, ‘Moonwalk’.

Here’s a look at some of the techniques being used to make it possible:

Temporal Recognition which analyses walking over a series of images. It uses techniques such as Gait Energy Images, Chrono Gait Images( CGI), Framework difference energy images ( FDEI), and gait entropy images ( GEnI)
Neural Neural networks including Convolutional neural networks, Deep Belief Networks, Recurrent Neural Network, Deep AutoEncoders, , Generative Adversarial Networks, Capsule Networks, Graph Convolutional Networks, 3D Convolutional Neural Networks. Basically the whole kitten caboodle is being thrown at the problem

Challenges faced by researchers:

camera view points
clothes on the person
occlusion
body part motion
large amounts of data processing
multi-task learning
ability to generalize unseen data
and the list goes on.

Who wants to bet that fashion will shift as a result of this technology being rolled out?

Read a survey of all the tech here.

Papers with code here.

Security

Machine Learning Security Operations ( MLSecOps) —a field I hadn’t heard about until last week— is fascinating. Apparently, those packages Data Scientists install to complete their work have security holes. Like any security hole, they leave information to hackers who know how to exploit it.

Many of the packages run via Application programming interface(API). When misconfigured they leave gaps that hackers can exploit.

On top of that AI powered hacking is on the rise.

According to the article, 51% of security leaders view AI-driven attacks as the most severe threat to their organization.
84% of IT and security leaders find that AI enhanced tools have exacerbated the challenge of detection phishing and smishing attacks.

Read full article here.

If you work in ML, I encourage you to read about it on this website.

And if you’re really interested read this article on Advanced API attacks in ChatGPT.

Interesting

How does one grow food in deserts?

This startup is finding out.

They developed heat-blocking roof and are working on resilient plant genetics. There’s a move to bring food closer to the consumer. It will be interesting to see how this shakes out.