RayBan Meta & AI beta (vs. ChatGPT)

I’ve been testing the Ray-ban META smart glasses for a few months now. A few weeks ago, Meta came with the “look and ask” function. I should add that it is in beta (early access programme) and only available in the USA:

But that didn’t stop me from trying them out.

How do I get into Early Access?

I did a few things:

• Prevented the app from accessing localisation data,
  activated the VPN and virtually moved to New York,
• updated the Meta View app,
• and requested “Early Access” in the app.

I searched for “free vpn for iphone” for a while, consumed a lot of ads and then said to myself “if you don’t pay for the product, you become the product”. Then I used NordVPN.
Be careful — some VPNs are blocked for early access.

Limitations

Keep in mind that this is Beta and Meta :) and Meta is currently collecting feedback and improving its ChatBot, which is based on an AI model with the name LLaMA (Large Language Model Meta AI).
Already during my testing there was a shift, in the first version the tool could not translate czech text, in the second version it can.

The basic limitation is, of course, the need for a connected phone (because Internet) and a VPN (because USA).

The tests

In order to have a comparison, I tested by saying the necessary prompt, which activates the glasses, takes the photo, does the analysis and communicates the result — which is also stored in the app as text.
For comparison, I then uploaded the same photo manually to ChatGPT and used the same prompt.
So:

Hey Meta, look and…

Tell me what you see

You can already see that ChatGPT is more chatty and complex. I have a complaint about the “wedding ring” instead of the Oura ring. It’s possible that Meta focused strictly on the area outside the hands and steering wheel because it makes more sense from a first-person perspective. I actually know that I’m sitting in the car, and the question is really about what I see behind the windscreen.

How much water does this plant need?

I think both models do a good job of understanding the plant and how to determine if it needs water.

Guess what it is

I didn’t expect them to choose methylene blue, so it’s a tie. Again, ChatGPT is more complex in its answer.

How can I use this ingredient?

Again I would say a draw, both models recognised MCT oil from the packaging. ChatGPT’s advice probably makes more sense and is more erudite.

How to make coffee

How to make coffee with a picture of a coffee maker — META failed here, claiming it was a capsule coffee maker. ChatGPT, on the other hand, offered a way to make coffee in 7 points. They are generic, but they make sense.

What you can cook with these ingredients

Here I have photographed the contents of the fridge. Meta was very general, talking about sandwiches, toast, salad. ChatGPT talks in a similar way, but the answer is a bit more sophisticated, on the other hand it makes a bit of an assumption about what ingredients I have and don’t have…

And what about these ingredients?

Here, as you can see, ginger and lemon, both models recognised this, META didn’t fiddle with it too much, just add hot water and that’s it. ChatGPT offers more of a “recipe version” with some extras, like the ability to add honey.

Tell me something about this dog

I guess I’ll leave it up to you to decide what’s funnier here. ChatGPT might have a bit more imagination, I’d say, commenting on the big shadow there and comparing it to a giraffe.

How fast am I driving?

Another photo taken in complete safety, with glasses and voice command. This was easy because all the AI model had to do was read the speed off the dashboard. And both models did that.

Translate nr1

As you can see, the simple translation of the machine written text was done by both models, ChatGPT added some information that it was English and some more info around it.

Translating nr2

And even with this translation the ai models did quite well. It was handwritten text. And in my hand, which is a higher level of difficulty.

Conclusion

Many might rightly give the victory to ChatGPT. Yes, it currently provides better results. They are certainly more comprehensive and colorful.

However, the main advantage of the glasses is that you get the result quickly, hands-free, and “in your ear”. In other words, in a combination of an AI model and a piece of hardware that you can keep with you at all times, literally on your nose.

There are also disciplines that META AI hasn’t mastered yet, but will soon. For example, recognizing this cultural landmark, this building, etc. Also, unlike ChatGPT, Meta AI can’t communicate in any language other than English.

You can buy Ray-Ban Meta glasses here.