A week in Generative AI: No slowing down, Zoom slips up and political biases uncovered
News for the week ending 13th August 2023
xxx
AI progress is unlikely to slow down
The chart says it all - AI is getting better at a faster rate than we’ve ever seen before. We’ll soon be measuring progress in weeks/months not years.
OpenAI announces GPTBot
GPTBot is a website scraper that OpenAI has just released. It will allow website to disallow it with a simple piece of code and allow OpenAI to ensure it removes sources that require paywall access.
Zoom's Updated Terms of Service Permit Training AI on User Content Without Opt-Out
This isn’t a good look for Zoom with many people becoming more privacy conscious with the rise of generative AI. Be interesting to see how long it will take them to back-track on this because I think they will…
Nvidia teams up with Hugging Face to offer cloud-based AI training
Great to see Nvidia supporting the largest, most important open source resource in generative AI.
New research visualizes the political bias of all major AI language models
Meta’s LLaMa model was found to be the most right-wing authoritarian, which is no surprise given the data it’s trained on. Interestingly the PhD researcher, Chan Park, believes that no language model can be completely free from political biases.
“The future is already here, it’s just not evenly distributed.“
William Gibson
This article was researched and written with help from ChatGPT, but was lovingly reviewed, edited and fine-tuned by a human.