First of all, happy new year, everyone.
Towards the end of last year, Google launched a "wow" A.I. demo using their latest technology called Gemini. It showed dazzling demos which allow users to input hand drawing, picture and text. Gemini will then interact with your request based on the picture. For instance, in the demo, you can show the picture of two yarns of different colors, and ask Gemini to suggest what you can knit. Gemini will show you the new knitwear. Neat.
Here is the link to the demo:
Some of our friends know that as a hobby, we like to do yoga from time to time. Yoga poses are well defined and an experienced yoga teacher can easily spot a mistake in a posture.
So, we thought, if we show Gemini a yoga pose that has a mistake, would it be able to spot it?
We went to a yoga web site: https://yogabullet.com/forward-fold/
which shows a few yoga correct and incorrect poses.
We then send the picture of the incorrect pose to Gemini. Guess what! It knows.
Here is the demo. This is what we enter, a picture and a text.
(In the following examples, text with purple blue background were entered by us --- yes the human. The text with the white background are from the computer --- yes the machine.)
The A.I. then spit out the following, which is spot-on.
Pretty amazing. But, we know we have friends who rightly have pointed out that we already told the A.I. that there is something wrong with this yoga pose. Kinda giving too much hint to the computer.
O.K. Let's try this one then. We take another pose. This time on how to hold a pair of chopsticks: https://chefjacooks.com/en/how-to-hold-chopsticks/
Here is our input:
With the following outcome.
It did not get the answer right the first time. However, with a little bit of nudging, it realizes that it has made a mistake.
In all the above examples, we did not do any fine tuning. It was created by just writing simple prompts.
P.S. You can try Gemini through Poe.com for free. All you need is to select the chat bot of Gemini like below. You can either click this link https://poe.com/explore?category=Official or click on the "Explore" on the left when you launch poe.com
P.P.S. Useless trivia: Poe's founder is on the board of OpenAI as well. During the saga of ousting the OpenAI founder (who later on returned to OpenAI), Reference: https://www.cnbc.com/2023/11/22/sam-altmans-back-heres-whos-on-the-new-openai-board-and-whos-out.html