Welcome to Nural's newsletter where you will find a compilation of articles, news and cool companies, all focusing on how AI is being used to tackle global grand challenges.
Our aim is to make sure that you are always up to date with the most important developments in this fast-moving field.
Packed inside we have
- GPT-4Chan: A large language model trained on hate speech and released on 4Chan
- Taser maker proposed shock drones for schools
- and DALL-E mini: translate any text into an image in 2 minutes
If you would like to support our continued work from £1 then click here!
Marcel Hedman
Key Recent Developments
Lessons from the GPT-4Chan Controversy
Andrey Kurenkov
What: "The worst AI ever". This is a quote about a language model named ‘GPT-4chan’, which was then deployed on the message board 4chan. The model was made by fine-tuning GPT-J with a previously published dataset to mimic the users of 4chan’s anonymous message board. The bots collectively wrote over 30,000 posts over the span of a few days, with 15,000 being posted over a span of 24 hours.
Key Takeaway: The work has been criticized by AI researchers, who claim the research would never pass human research ethics boards. To which the author responded that the model isn't doing anything existing large language models are not capable of. What can and should be done in the name of "research"?
Taser maker proposed shock drones for schools. What could go wrong?
What: In response to the horrific Uvalde school massacre, police contracting firm Axon proposed flying Taser drones in school ceilings that could be launched remotely by police officers. The response from the Axon board was very clear, with nine members resigning following the proposal. The resignations were out of fear the drones would stun innocent students or be abused by hackers, vandals or the police.
Takeaway: The use of ML and drones to tackle some of the largest challenges are growing. However, questions must always be asked of the proposed implementations and considerations taken for the potential adverse externalities that can follow.
Dall-E mini: how the AI image generator makes your meme dreams come true
What: Many have come across DALL-E mini, an ML-based image generator which turns input text into images. Almost anything can be generated using the language model with some very interesting results.
Takeaway: Some have expressed how the model augments some existing biases within society, as the model is a reflection of the bias inherent within the training data. On a lighter note: others are using the model for social media content and memes!
AI Ethics
🚀 Looking for a connection in AI: fanciful or natural?
🚀 The Imitation of Consciousness: On the Present and Future of Natural Language Processing
🚀 Facebook Is Receiving Sensitive Medical Information from Hospital Websites
🚀 That AI Image Generator Is Spitting Out Some Awfully Racist Stuff
Other interesting reads
🚀 AMA: I left Google AI research after 3y
🚀 It's Not All Fun and Games: How DeepMind Unlocks Medicine's Secrets
🚀 Facebook AI Research goes through massive restructuring
🚀 Microsoft release PyWhy: open source home for causal inference projects
🚀The Ethical AI Startup Ecosystem 02: Data for AI
Cool companies found this week
Health
Aidoc - Maker of AI-based software that helps radiologists read medical scans and alerts them to strokes or pulmonary embolisms, has pulled in a $110 million Series D investment round.
Research
Cohere For AI - A non-profit research lab that seeks to solve complex machine learning problems. They support fundamental research that explores the unknown, and are focused on creating more points of entry into machine learning research.
...and Finally
AI/ML must knows
Foundation Models - any model trained on broad data at scale that can be fine-tuned to a wide range of downstream tasks. Examples include BERT and GPT-3. (See also Transfer Learning)
Few shot learning - Supervised learning using only a small dataset to master the task.
Transfer Learning - Reusing parts or all of a model designed for one task on a new task with the aim of reducing training time and improving performance.
Generative adversarial network - Generative models that create new data instances that resemble your training data. They can be used to generate fake images.
Deep Learning - Deep learning is a form of machine learning based on artificial neural networks.
Best,
Marcel Hedman
Nural Research Founder
www.nural.cc
If this has been interesting, share it with a friend who will find it equally valuable. If you are not already a subscriber, then subscribe here.
If you are enjoying this content and would like to support the work financially then you can amend your plan here from £1/month!