#News

The Week in AI - Italy’s DPA Blocks ChatGPT-4, AI Detects Stress Deformation in Brain, and More!

The Week in AI - Italy’s DPA Blocks ChatGPT-4, AI Detects Stress Deformation in Brain, and More!

Date: April 10, 2023

A lot has happened in the last week of AI. From ChatGPT’s temporary ban in Italy to AI being used for understanding deformations of stress in the brain, here are some of the most interesting news related to AI and machine learning below.

A lot has been going on in the world of the AI industry. In recent news, the Data Protection Authority (DPA) of Italy has temporarily blocked ChatGPT. It is under concerns that ChatGPT is violating GDPR (General Data Protection Regulation) laws. Generative AI tools like ChatGPT scrap data from the internet to train their model and provide results. These tools also use personal user data from Twitter handles and Reddit without even notifying the user leading them to violate the law.

OpenAI, the company backing ChatGPT, must respond within 20 days to settle the legal fight. Privacy policies such as GDPR are surely gonna open up Pandora’s box. The fight may have started in Italy but its implications are more likely to create a dialogue about privacy concerns around the world.

In other news:

  • Microsoft to Explore the Possibility of Using Ads in Bing Chat: Microsoft has been experimenting with Bing Chat using GPT-4 generative AI language model. This could be a new wave of advertising, however, this could decrease the trust in the tech considering it already struggles with factual errors.
  • Temporary Pause on training AI systems: A total of 1100 signatories including Elon Musk, on Tuesday, signed a letter to halt the training of AI systems more powerful than ChatGPT for the next 6 months. However, a majority of signatories have backed off in the subsequent days, and there have reportedly been signatories that turned out to be fake.
  • Twitter makes its Algorithm Public: As promised by the current CEO of Twitter, Elon Musk, a small portion of the Twitter algorithm’s source code has been made open source. The algorithm seems to use a neural network to rank content based on likes and replies. 
  • AI-based Summarizing Tool: A meeting intelligence tool similar to Otter and Zoom with name “Read”, has released a new feature that can trim hour-long meetings into minutes. The company is using language models and video analysis to pick up important parts. 

More News on Machine Learning…

A lot has taken place in the last week of machine learning updates:

  • AI Enabler Nvidia Fine Tune’s Biotech Platform: A major hiccup in the AI tech industry is processing huge heaps of data for churning out information. Nvidia’s new strategy BioNeMo has undergone a few changes that have made the platform more accessible to users. The intention behind the entire project is to improve operational efficiency in the research department. With Nvidia’s acceleration, the aim is to achieve efficiency in doing the same work for up to 5 to 10 projects instead of just one.
  • Data Journalist Investigates Medical AI Process: After a year-long battle with breast cancer, Meredith Broussard goes on to investigate the role AI plays in the medical processing of mammograms. After multiple trials and attempts to find the truth, she came to the conclusion that taking ownership of the data and processing is complicated for a patient.
  • Persuasion Capabilities of Tools like ChatGPTStandford conducted a new study where they tried to understand the persuasion capability of tools like ChatGPT. The study found that content generated by ChatGPT was as persuasive as its human counterparts. In fact, the tools may have the capacity to persuade in a political context too.
  • New Insights in Brain Injury Via AI: Artificial Intelligence has successfully decoded the mystery behind how stresses reflect in forms of deformation inside the human brain. The initial issue was that there was no fixed pattern in the brain that reflected similar stress issues. It is because unlike many things human brain is more like jello-o i.e. it's not homogeneous. The AI model used has the capacity to use previous computational models and combine them to provide insights. This is carried out via mix-and-match.
Arpit Dubey

By Arpit Dubey LinkedIn Icon

Arpit is a dreamer, wanderer, and a tech nerd who loves to jot down tech musings and updates. With a logician mind, he is always chasing sunrises and tech advancements while secretly preparing for the robot uprising.

Have newsworthy information in tech we can share with our community?

Post Project Image