aifaq.wtf

It's a little in the weeds of how these tools work, but generative AI tools are based on "tokens," individual units of text or information. Usually the tokens are developed organically by the model itself, but in this case, the tokens were developed by the team at Baichuan, and they're very revealing of the model's... biases? I don't think I'd call them a bias, just... a reflection... of... something.

Read the responses.

Running models locally is wild! I've seen people talk about it plenty before, but the video really sold it to me.

I need to go buy a fancier mac! I've always prided myself on using old cheap Airs buuuuut maybe that era is over.

Meta released a new model that can do all sorts of speech- and text-based communication tasks. It's called Seamless, and you can find the model here.

Speech-to-speech translation is super fun, yes, but I love that they highlight that it (supposedly) understands code switching, aka bouncing back and forth between different languages. That's always been a tough one!

I tried the demo here and mostly learned my Russian is horrible:

Tried again with my pathetic grasp Japanese, and it thinks I'm speaking Hindi. I feel like these systems are usually pretty good at coping with my inability to speak other languages well (or at least Whisper is), so I don't know if I'm especially bad today or this model's flexibilty is also a potential downside.

All you need is a few tiny interconnected pieces and you get a mean ol' robot.

All of the zinger-length "AI did something bad" examples used to go under #lol but this is getting more and more uncomfortable.

Courtesy the AI Snake Oil newsletter.

I thought journalism was the part of AI that was most dangerous to have generative AI errors creeping into, but I was wrong.

Wow, this is not a joke. Byline is "Story by Microsoft Travel," and the Ottawa Food Bank is definitely number three.

3. Ottawa Food Bank

The Ottawa Food Bank gives fresh and non-perishable food, as well as diapers and other necessities.

The organization has been collecting, purchasing, producing, and delivering food to needy people and families in the Ottawa area since 1984. We observe how hunger impacts men, women, and children on a daily basis, and how it may be a barrier to achievement. People who come to us have jobs and families to support, as well as expenses to pay. Life is already difficult enough. Consider going into it on an empty stomach.

This tweet has a screenshot if it disappears.

Okay so no feathers but this is perfect:

In this work, we introduce a new task Skull2Animal for translating between skulls and living animals.

You thought that was it? Oh no no no – the deeper you go, the more morbid it gets:

With the skull images collected, corresponding living animals needed to be collected. The Skull2Animal dataset consists of 4 different types of mammals: dogs (Skull2Dog), cats (Skull2Cat), leopards (Skull2Leopard), and foxes (Skull2Fox). The living animals of the dataset are sampled from the AFHQ.

Skulls aside, this is to be my go-to example of AI perpetuating culturally-ingrained bias. Dinosaurs look like this because we say dinosaurs look like this. Jurassic Park was a lie!

While attempting to find problematic dino representations from Jurassic Park beyond lack of feathers, I came across a BuzzFeed list of 14 Times Jurassic Park Lied To You About Dinosaurs. It's actually a really good, well-sourced list. I think you'll learn a lot by reading it.

Paper on arxiv, which seems to actually be all about dogs despite the title. Here's some more nightmare fuel as compensation:

Filing this one under AI ethics, bias, and... media literacy?

Reminds me of the paper by Anthropic that measured model bias vs cultural bias.

AI detectors are all garbage, but add this one to the pile of reasons not to use them.