Elon, LLaMA, and CyberHackers

Alright. Here we have it. Lot’s of stuff happening.

So before we thread the needle on these topics, lets have a post-mortem for threads:

Well, seems like Elon’s in the lead once more. Once their web app comes out, could it even out?

Who knows, but keep your eyes peeled.

Speaking of Elon, he announced XAI this week, a company dedicated to developing an AGI.

Welp, we’ve got another player in the LLM race. So many to keep up with -OpenAI, Anthropic, Google, open source ones like HuggingFace, and now…

XAI.

The twist? It will work with tesla for its data integration capabilities. Actually, not much of a surprise…

What is interesting, though, is that Twitter data will help train the model and its generative capabilities.

This is very reasonable, and considering how other LLMs scraped twitter as training data for their own LLMs

*Cough OpenAI Cough

Its like the AI avengers…

Anyways, moving on.

Wimbledon

You’ve ever heard of Wimbledon?

Well, they’ve been implementing AI into their commentary for highlights. It’s a smart move from them to expand bandwidth, but it’s been getting some pushback from tennis fans.

One commentator said it’s “stiff, emotionless, and can kill humanity”. Quite dramatic, but many do agree.

And anyways, the tech is out there– the genie is out of the bottle. You’ve just gotta use it

Worm GPT

Now, this has got nothing to do with worms, or ChatGPT for that matter.

Then, what is WormGPT?

Well, it is a spinoff of ChatGPT based on the 2021 GPTj model, that- get this– hackers are using to commit cybersecurity attacks.

Interestingly enough, if a developer looks at the tool’s functionality, without concern about its use- it seems wonderful. Chat memory retention. Code-formatting chops. Unlimited characters.

It’s got the whole 9 yards. The whole sha-bang.

The catch?

It’s been trained on malicious and compromised datasets. WormGPT, which does not have the safety measures in place that ChatGPT does, is being used to execute phishing attacks on business emails.

And so, regulators might use this as the linchpin to squash the acceleration of AI development. But…there’s a line between stopping innovations and stopping malicious actors.

And its gotta be walked.

LLAMA 2.0

Now for Llamas- not those kind. We’re talking about the open source LLaMA 2 model that Meta released yesterday.

It’s an open-sourced model that is jam packed with licenses and agreements- intended for commercial use.

Some call it a nuked model because of its extensive safety evaluations and alignment…but, others are saying it outperforms GPT-3 on many tasks.

Hey, not bad.

Microsoft also invested in it- Satya’s really investing into all the AI giants- OpenAI, now LlaMA. Dude obviously knows what he’s doing.

Anyway, we’ll watch how all this unfolds.

Till next time.