Google Admits Biased Model Flaws In Gemini

Political parties are going all out with AI this Election Season

Sponsored by

We explain the latest business, finance, and tech news with visuals and data. 📊

All in one free newsletter that takes < 5 minutes to read. 🗞

Save time and become more informed today.👇


Hola!! Creators,

Google’s recent Gemini(1M token context) release was found to be a tad too woke when asked to create certain images, guess that’s why Google has a $60M deal in place with Reddit to buy user-generated content from Reddit to train its AI model. Learn the latest Principles of prompting techniques as we see the application of Generative AI in different fields.

Today’s Agenda

  1. Generative AI In Different Industries And Their Applications

  2. 5 Principled Prompts For More Efficient GPT4(LLM) Responses

  3. What To Believe In The Age Of Political Deepfakes ?

  4. Google Puts A Hold On Its Woke Image Generator In Gemini

  5. Sora : Text-To-Video Tool By OpenAI Leaves Everyone Stupefied

  6. Prompts For Generating Distortion Free Faces on SDXL

Newsletters You Should Check Out

  1. AI For Normal People : Use AI in your normal life and harness the maximum out of it- Read

  2. Simple.AI : Simplifying AI for consumption and usage as per your requirement:- Read

  3. AI Hustl: AI for the hustlers who want to stay ahead of the GenAI trends and learning curve- Read

  4. The Creator Lens : Another one for the Creators, stay on top of the trends and current landscape- Read

Generative AI In Different Industries And Their Applications

5 Principled Prompts For More Efficient GPT4(LLM) Responses

As we make advances in mastering the art of generative ai and speak the language of LLMs by prompting, let’s take a look at a few recently researched principles which ensure a more effective and efficient solution from the LLMs. These principled prompts are 26 in numbers and can be accessed here

  1. Structured Prompts: Provide clarity to the LLMs around the tasks, structure the information for more effective comprehension. Integrate the intended audience, have affirmative directives like “do” while staying clear of negative like “don’t”

  2. Specific but Informative: Being as specific as possible can be highly effective. Implement an example-driven prompting style where you provide a few examples for ideal solutions/approach. Specify in prompt to “ensure that response is unbiased”.

  3. Interact & Engage : Allow the model to elicit precise details and requirements from you by asking questions until it has enough information to provide the needed output. Respond with positive or negative affirmations to give more information about the quality.

  4. Content & Language Style: Include phrases like “your task is”, “You must”, “you will be penalized if..” to get more accurate results and responses. Assign a role to the language model first so as to provide complete clarity. repeat specific word multiple times.

  5. Complex Task Breakdown: Break down the complex tasks into a sequence of simpler prompts in an interactive conversation, combine chain of thoughts prompts with few-shot prompting for getting more accurate results for coding tasks as well.

What To Believe In The Age Of Political Deepfakes

Political Disinformation campaigns have existed since long but AI has democratized disinformation by making it easier even cheaper in producing quick forgeries. Be it the faked robocall apparently from Biden urging Democrats not to Vote in New Hampshire or Sadiq Khan, the mayor of London calling out the faked audio which supposedly captures him making incendiary remarks about Remembrance weekend. The issue is pertinent and even more worrisome now since nearly 40 nations worldwide are set to enter into election modes soon. Know More

Google Puts A Hold On Its Woke Gemini

The Supposed Alternative To ChatGPT-Gemini with 1M Token Context Is WOKE

Google recently rebranded its Generative AI to Gemini and its not even days since its already in the news. While Gemini currently boasts of the ability have the capacity of 1M token context, which is much larger than that of all its competitors’ combined. However that is not the reason for Google being in the news since now Gemini is being termed as being too “WOKE” by the users as Gemini showed inaccurate images when asked to furnish information about the Founding Fathers of the United States. Google admits to the flaw and has paused the image generations till it gets fixed. Read more

Google however has also been in news for a $60M deal with reddit to utilize user generated content as datasets for its Generative AI. Admitting to the large number of reddit specific queries received by the search engine, it makes sense to train the models on human content so as to make it sound more like human while ensuring it doesn’t stick to a bias like being woke in generating images from history. Know More

Sora : Text-To-Video Tool By OpenAI Leaves Everyone Stupefied

Now with just a few keystrokes, get ready to create exceptional videos with OpenAI’s text-to-video generative AI model which can simply turn short text prompts into strikingly realistic videos up to a minute long. The creations shared so far have evoked a mixed reaction but the general consensus is everyone is stunned at the results. However fear being one of the major emotions, people are not shying away from expressing the same openly around the immense and rapid development of technology and whether its desirable or not. Read More 

Prompts For Generating Distortion Free Faces on SDXL

Check out these distortion free face images created on SDXL

Reply

or to participate.