OxJosh

Technical Crypto Writer | Web3 Developer Content Marketer

[OxJosh @OxJoshWeb3]

and more inclusive perspectives shape the future of AI.

What do you think? Is coordinated pausing even realistic at all?

#LLM #AI #ResponsibleAI #AISafety #AIregulation #coordinatedpausing #machinelearning #AIAlignment #AIGovernance

[OxJosh @OxJoshWeb3]

AI is getting way more powerful and building its very own progenitors...

Anthropic just highlighted what many in AI governance have been warning about, the path to recursive self-improvement and the urgent need for coordinated global mechanisms to manage it.

A 🧵(1)

[OxJosh @OxJoshWeb3]

Unilateral restraint is unrealistic, but neither is racing blindly.

As an African and AI safety enthusiast, I believe we need more representations from the Global South in these governance conversations in order to reduce the digital divide and ensure diverse...

(8).

[OxJosh @OxJoshWeb3]

(6). Your model might be a superstar in your local environment, trained and tested on carefully made data but the real world is messy and unpredictable.

And in high-stakes domains like healthcare, that gap isn’t just a performance issue,
it’s a matter of life and death.

[OxJosh @OxJoshWeb3]

(1). I once built a convolutional neural network to detect and classify brain tumors using MRI images and it nearly cost a patient her life...

Here is how it happened...

[OxJosh @OxJoshWeb3]

(5). No matter how much augmentation you apply, you’re still simulating reality, not capturing it.

So the model learns patterns that work beautifully in a sandbox…
but fail dangerously in the wild. That experience changed how I think about machine learning systems.

[OxJosh @OxJoshWeb3]

(4). That’s when the real issue became clear. The problem wasn’t just the model, it was the data.

Many publicly available datasets (like Kaggle) are collected in controlled environments. They are clean, balanced, and often lack the variability that define real clinical settings.

[OxJosh @OxJoshWeb3]

Your antivirus is probably blind.

Not broken. Just… outdated.

Because today’s attackers don’t match signatures.

I broke down why signature-based security is failing and how UEBA uses machine learning to catch threats hiding in plain sight.

[https://t.co/SFH2c6S0U4]

[OxJosh @OxJoshWeb3]

@smartnakamoura Me: npm install
Random person: "He is an hacker"
😂😂😂

[Samantha Simonhoff @RealProductGirl]

I NEED my feed full of builders.

What are you working on right now?

I don't care if it's a startup or a weekend side project. If you're building something, I want you on my timeline.

Reply and let's connect. 👇

[OxJosh @OxJoshWeb3]

Been a bit quiet lately was quite intentional.

The past couple of months have been about building, learning, and going deeper.

Lately, I’ve been exploring the intersection of cyber security and AI and I’ll be sharing more of that here soon.

Back to building. ⚒️✨
#ML #AI