🤖 GigaChad or ChadGPT?

What’s good?

We’re back with your daily dose of AI.

Today we’re Black Mirror themed with GigaChat, Robo-Dogs and questioning this all through the lens of ethics.

🗞 News in 30 secs

1/ Russia launches GigaChat

Financial services company Sberbank has OpenAI competitor GigaChat. It’s a neural network trained and optimised for Russian. It’s currently invite-only so we’re yet to see how it stacks up.

2/ Boston Dynamics add ChatGPT to their Robot Dog

Remember that episode of Black Mirror where the robot dog chases the protagonist for 41 minutes? Well… we’re basically there. It’s like we’ve learned nothing. Funnily enough, Boston Dynamics’ mission is “to imagine and create exceptional robots that enrich people’s lives” which begs the question How does a robotic dog enrich my life? What are the use cases and why does it need to be more OP? Watch the BM episode to find out.

3/ NVIDIA releases NeMo Guardrails

An open source project aimed at helping developers ensure the security, accuracy, and appropriateness of AI applications using large language models (LLMs) - it prevents AI from hallucinating wrong facts. Check out the project here.

Ethics…. Who needs em’?

So as we continue to integrate AI into our lives and give it more direction over our decision making, we need to be more aware of the ‘thing’ making that decision or writing that content.

Tucker Carlson’s back on the market. But I doubt he’s going to get a job at buzz feed.

Why?

Because BuzzFeed terrible executives have access to Carlson’s terrible record and can see that whilst equally terrible they are at opposing ends of the political spectrum.

For this same reason, we must be aware of the bias & ethics of this beast that has now been unleashed to make decisions for us.

Algorithmic bias. It’s the term used to describe AI systems that seem to discriminate between certain groups of people or lean to a certain side of the political spectrum.

Some examples include not arguing for fossil fuels or only wanting to write poems about Joe Biden and NEVERTRUMP. (Despite Trump being super bullish on GPT)

In fact, ChatGPT was put through a full political spectrum assessment and found to be left leaning..shock.

That’s not all though. It gets consequential fast

1/ Some facial recognition algorithms are less accurate with darker skin tones and women.

2/ Criminal justice algorithms have been shown to exhibit racial bias.

3/ Some hiring algorithms have been found to discriminate against certain groups, such as women.

4/ Language models can exhibit biases based on the language used in the training data.

We explained why this happens in our article about Neural networks. AI models are only as good as the data you give it.

Feed it crap. Get crap.

Or more specifically, feed it left wing data get left wing responses etc etc.

So what’s being done about it?

Well? A few interesting things.

1/ Companies like OpenAI want to set up community forums. The purpose of these forums would be to collectively decide what should be outright banned. (Sounds pretty black mirror too…)

2/ A Customised Chatbot is coming. Open AI are also working on a bot that can be tailored to have the same world view as the user. e.g they will train their AI models to represent different perspectives and world views.

Until then…

You can jailbreak pretty much any GPT model 👀 

But that’s for tomorrow.

What’s Appnin’ 🎮

RunwayML launches native Iphone app Gen-1.

It is a creative suite of AI tools and allows artists to create media ranging from videos, images to audio all with text.

Basically, if you can dream it, you can now create it.

Creators are no longer limited by their Photoshop or DaVinci Resolve skills.

You’ll only be limited by your imagination.

Below is from a reddit post showing the speed at which AI art is moving (by Midjourney a RunwayML competitor).

Human or Not?

With AI getting more and more sophisticated, it’s becoming increasingly harder to discern whether something was written by a human or an AI.

This game puts this to the test, where you’re paired up with

someone… or

something…

by chatting to them you have to decipher whether the entity at the other end is human or artificial intelligence.

Here is my attempt at bot hunting