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.
Last week marked the close of Graham Lane's time as weekly newsletter writer at Nural. Graham, your impact on the Nural community will not go missed and we look forward to keeping in touch as you continue to make your mark on the tech world!
Packed inside this week's newsletter we have
- Image-Generating AI Keeps Doing Weird Stuff We Don't Understand
- An AI shopping app that was actually just using humans
- and will a fruit-picking robot transform agriculture?
If you would like to support our continued work from £1 then click here!
Marcel Hedman
Key Recent Developments
Image-Generating AI Keeps Doing Weird Stuff We Don't Understand
What: A viral Twitter thread (and accompanying un-peer reviewed paper) has claimed DALL-E, the popular text-to-image generation system created by OpenAI, is creating its own language of gibberish words and matching them to classes of images like birds and insects.
Key Takeaway: Understanding the inner workings of large language models is still an open topic. Is the case described above a potential peek into this confusing space or just a case of extreme confirmation bias...
AI-driven robot boat Mayflower crosses Atlantic Ocean
What: A crewless ship designed to recreate the Mayflower's historic journey across the Atlantic 400 years ago has completed a 2,700-mile (4,400km) trip from Plymouth in the UK to Halifax in Nova Scotia, Canada. Despite not making it to its intended destination of Massachusetts, this represents a significant accomplishment for the unmanned vessel.
Key takeaway: The ship was navigated by on-board artificial intelligence (AI) created by IBM with information from six cameras and 50 sensors. As well as representing a fun recreation of a historic journey, the trip has created invaluable data and learnings about maritime enterprise.
An AI shopping app that was actually just using humans
What: A company which billed itself as an “artificial intelligence startup” that uses AI to auto-fill customer information for $1 per transaction, has been reported to be using humans to manually handle the form completion for 60-100% of instances in 2021. The app would reportedly save shoppers a few minutes when completing purchases via the app by completing.
"People with direct knowledge of the technology used reported bot blockers on retailer sites as a problem. This resulted in a large bulk of transactions going through manual entry by actual humans. Some orders were placed hours after users pressed the “buy” button,"
Key takeaway: The line between marketing and lying is a well trodden one. AI systems often take a human-in-the-loop approach to ensure quality and safety, but for investors expecting to be financing a new state of the art technology, this approach may often be misaligned to what they were anticipating.
AI 4 social good
🚀 Artificial intelligence tool learns “song of the reef” to determine ecosystem health
🚀 Combating climate change and grid energy - the role of AI
🚀 Will this fruit-picking robot transform agriculture?
Other interesting reads
🚀 Lack of data compromises research on climate change
🚀 Real-Time AI Model Aims to Help Protect the Great Barrier Reef
🚀 Helsinki’s pioneering city digital twin
🚀 Dear Elon Musk, here are five things you might want to consider about AGI
Cool companies found this week
Banking
Engine by Starling - Engine is banking as a service, taking all of the banking infrastructure under the hood for those looking to launch a bank in new regions.
Health
Insilico Medicine - End-to-end, artificial intelligence-driven pharma-technology company with a mission to accelerate drug discovery and development by leveraging their proprietary platform across biology, chemistry and clinical development. Recently raised $60m in series D funding.
...and finally
Fruit picking robot
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!