š Google's DeepMind Enters the GenAI Game
š Google's DeepMind Enters the GenAI Game
and ChatGPT is apparently left wing
Good morning, everyone! āļø
While this week has been relatively quiet, the bright side is that the UK weather has picked up. āļø
Anyway, let's dive into this week's edition! š
Whatās Been Cooking
DeepMind Building āLife Adviceā AI š§
Googleās DeepMind, an AI company most known for developing AI systems such asĀ AlphaGoĀ andĀ AlphaFold, has been building 21 AI tools for personal development and professional use.
The concept behind these tools is to provide advice on different situations that can occur throughout our lives. For example, you may ask about the most appropriate way to ask for a pay raise at work.
However, this once again leads to the āAI control our lives/ territory and how humans interact with chatbots. Experts are concerned with the ethical implications of the relationships we have with this technology.
ChatGPT Exhibits Political Bias š¤
According to aĀ study from the University of East Anglia, ChatGPTās responses are biased toward left-wing politics, favoring the UKās Labour Party and Americaās Democrats.Ā Elon Musk has previously declared biases, but this is the first academic study claiming this.
Lead researcher Dr. Fabio Motoki said that āThe presence of political bias can influence user views and has potential implications for political and electoral processes.ā
The test for the bias came from asking the chatbot to assume the roles of different individuals throughout the political spectrum and ask 60 political questions. These were then compared to the regular state of the chatbot to determine where on the political spectrum its affinity lies.
Side Dishes
Some more tasty stories this week:
TUI Has An AI Tour GuideĀ āļø
Insider Insights
Make and Makefiles In Machine Learning PipelinesĀ š„ļø
Nowadays, Data Scientists are using software engineering practices more than ever. The Linux āmakeā command is popular with engineers, however, it is relatively unknown among Data Scientists. The āmakeā command allows you to build and break down your modeling pipeline for easier debugging.Ā
Reading & Listening
A couple of blogs that caught my attention this week!
NP-What? Complexity Types of Optimization Problems Explained: A great blog post detailing the time complexity of optimisation algorithms, with fantastic visuals.
Why More Is More (in Artificial Intelligence): An interesting exploration of stochastic gradient descent and the double descent phenomenon for neural networks.