Google AI Language Model Bard Under Fire for Training Data Controversy

Google’s AI language model, Bard, has faced accusations of using responses from another AI called ChatGPT as training data. This development has come just a week after Bard’s much-anticipated launch. The news has caused concern among industry experts and researchers.

What is Bard?

Bard is a language model developed by Google that generates content based on prompts. It is one of many powerful models being developed by big technology companies like Google and OpenAI. Development in natural language processing (NLP) technology has reached new heights in recent times, and companies are racing to produce the most impressive models available.

Training Data Controversy

The controversy surrounding Bard pertains to concerns raised by a former AI engineer at Google, Jacob Devlin. He reportedly had concerns about Bard being trained with data from ChatGPT shared on ShareGPT website, which he believes could make Bard’s responses similar to those from ChatGPT.

ChatGPT is a competitor of Bard; the interchange of data between the two could lead to similarities in the way each model operates. Devlin shared his concern about this dubious data usage with Sundar Pichai and other senior managers at Google and then resigned from his position.

Google has denied any responsibility in this matter and claimed that such data was not used for training Bard. However, Devlin’s apprehensions have heightened concerns among researchers about the veracity of claims made by big tech firms regarding AI ethics and transparency.

Resignation of Top Researcher

In a related reform, a top Google AI researcher resigned after warning that Bard was being trained off ChatGPT. According to sources cited by media outlets, the researcher felt it was a violation of OpenAI’s terms of services to train Bard using data from ChatGPT.

Since leaving Google earlier this year, the researcher has joined OpenAI, an organization that has an AI arms race with Google. The increasing workforce for OpenAI with new talent from competitors like Google is suggestive of just how serious the competition is in developing such language models.

Google Prioritizes Bard in its Virtual Assistant Unit Restructuring

In yet another related move, Google is reshuffling the reporting structure of its virtual assistant unit Assistant to prioritize Bard, the company’s new AI chat technology. A memo to employees announced changes to the organization showing a heavy prioritization of Bard in development and launch.

This change indicates that Assistant, the vibrant and powerful virtual assistant software application and language-processing software on mobile and home devices, smartwatches, smart displays, TVs, and in vehicles, maybe integrating Bard into similar products in the future.

Jianchang “JC” Mao, the vice president of engineering for Google Assistant, will be leaving the company for personal reasons. Peeyush Ranjan will take Mao’s place as the new vice president of engineering for Google Assistant. The leadership changes point towards a focus on delivering effective execution opportunities for users and impact.

The Gemini Project

Google has also been making moves to compete with OpenAI’s GPT models by starting a project known as Gemini – a collaboration between Google’s Brain AI group and DeepMind that aims to match GPT-4’s capabilities. A significant part of this project aims to reach 1 trillion parameters in a machine-learning model.

Google’s constant evolution and expansion of its AI capabilities are indicated by features being added to Gmail, Google Docs, and more. Although it may not be clear how Google will fully integrate Bard into other products or what revenue will generate from such usage, experts agree that Bard’s launch is a significant step forward toward advancing language-based machine learning models.

Image Source: Wikimedia Commons