Gemini 1.5 Pro

Hello everyone,

Today I want to share with you an insane technology. Google recently released Gemini 1.5 Pro and it’s SUPER powerful. In this post, I will explain what Gemini 1.5 Pro is capable of and what features it brings. 

The first big feature is the input size. Gemini 1.5 Pro has a GIANT input capacity of 1 Million Tokens. That's about 700,000 Words. In comparison to all the other AI Chatbots, Gemini just puts them to shame.

Words aren't the only thing that Gemini can process. It’s also capable of taking a 1-hour-long video, 11 hours of audio and more than 30,000 lines of code as input. As you can see on the picture, Google even managed to bring it up to 10 Million Tokens in research. The fact that Gemini can process 1 Million Tokens is already insane and super revolutionary, but just imagine if Google decides to release the 10 Million version. Companies like OpenAI etc. have to catch up really fast. OpenAI has an excuse, though, because they have been working on the video generation model Sora for some time. But now they have to try to get GPT-5 going, and it has to be really good. Otherwise, Google will take over the Chatbot space easily. 

Another crazy thing is that, even with such a large Token capability, Gemini is extremely accurate. This was tested with the “Needle in a Haystack” (NIAH) problem, and it works like this: Gemini is given a 1 Million token long text and in this text, there is a small piece of text with a statement or a fact. Gemini was able to find that small piece of text 99% of the time. In a text that is 700,000 Words long, this is very impressive, and it shows that Gemini can be very accurate even with such a large input. 

Gemini also shows very impressive “in-context learning” capabilities. That means that it can learn a new skill just by giving it a long prompt explaining the thing it should learn. Gemini was tested on this Experiment: Gemini was given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person learning from the same content.

Sadly only a select few can access Gemini 1.5 Pro right now. We probably have to wait a bit for the full release.

These are all very impressive feats by Google, and I'm excited to find out, what Google can do in the future. 

Have a nice day and I’ll see you next Friday.

PS: I’ve been thinking about also posting a Newsletter on Tuesday. Would that be something that you’re interested in? Just let me know by replying to this email.