What's New In ML #2
World's longest drone superhighway, AI in football, Meta AI's BlenderBot 3, building 3D objects from 2D instructions, 10,000x faster artificial synapses, startup funding
Hello readers,
Welcome to What’s New In ML. The goal of this segment is to help you stay updated on what’s happening in the world of ML, AI, and Data. We’ll talk about real-world applications, cutting edge research, startups, and everything else that’s moving this field forward.
If you have a question, submit it here and I’ll get back to you with my thoughts. Subscribe to this newsletter to receive it in your inbox every week:
The biggest news in AI this week is that I got access to OpenAI's DALL-E system. It's an AI system that can generate fantastical images based on text descriptions. I'll let my creative juices flow and see what comes out.
Now that we got the most important news out of the way, let's see what else is happening in AI.
APPLICATIONS
UK Government gives the green light for world's longest drone superhighway
UK announced that it has given the go-ahead to for the world's largest and longest network of drone superhighways to be built within the country. It will enable pilotless drones to be flown beyond-visual-line-of-sight (BVLOS). Read more
Champions League to use AI to help detect offside
AI is finally coming to football (I refuse to call it soccer). UEFA has announced that it will start using the new Semi-Automated Offside Technology (SAOT) from the Champions League group stage this season. Players pretending to complain that they were never offside hopefully won't complain anymore. Read more
Meta AI releases BlenderBot 3
BlenderBot 3 is the first 175B-parameter, publicly available chatbot. They've provided the model weights, code, datasets, and model cards. They've also deployed it in a live interactive demo here https://blenderbot.ai/. It's capable of chatting about any topic. It searches the internet to discuss a particular topic and it's designed to learn how to improve its skills. I chatted with the bot to see how it responds. It's rough around the edges, but at least they're attempting to build this in public. It does well on simple conversations. But it's difficult to keep a human engaged when they choose random topic. Read more
Mount Sinai unveils first-of-its-kind department to develop AI tools for healthcare
Mount Sinai opened the Department of Artificial Intelligence and Human Health this week within its Icahn School of Medicine. It's the first of its kind to be launched at a US medical school. The goal is to develop decision-support tools driven by AI that will help physicians deliver hyper-personalized care to each patient. Read more
RESEARCH
Carnegie Mellon University open-sourced auton-survival
CMU has released auton-survival, an open source package for regression, counterfactual estimation, evaluation, and phenotyping censored time-to-event data. Making decisions in the real world requires reasoning about when an event will occur. The goal of this reasoning is to help with prioritization and intervention at the right time. It has applications in healthcare, bioinformatics, ecommerce, finance, and more. Read more
Researchers develop an AI system to build 3D objects using 2D instructions
Researchers at Stanford, MIT, and Autodesk have collaborated to develop a new AI system that can interpret 2D instructions to build 3D objects. This system is called MEPNet (Manual-to-Executable-Plan Network) and it was tested on computer-generated Lego sets, real Lego set instructions, and Minecraft-style voxel building plans. They said it outperformed existing methods. They also said their aim is to create machines that help people assemble complex objects, including furniture. Read more
MIT researchers train an ML model to control digital manufacturing
They have built an ML model to monitor and adjust the 3D printing process to correct errors in real-time. The system uses computer vision to watch the manufacturing process and then correct errors in how it handles the material in real-time. Read more
Artificial synapses that are 10,000 times faster than biological ones
Researchers have experimented with artificial neurons and synapses. These are one-thousandth the size of biological neurons and 10,000 times as fast as biological synapses. They mention that the speed of thought in animals is limited to milliseconds. It's constrained by the weak voltages and watery medium in which neural signals are shuffled. But solid-state neurons and synapses don't have these constraints. In this study, researchers experimented with artificial synapses called programmable resistors. They are essentially electric switches that can remember which state they were toggled to after their power is turned off. Read more
New ML algorithm aces university-level mathematics questions
MIT researchers use ML to automatically solve, explain, and generate university-level mathematics problems at a human level. Read more
STARTUP FUNDING
Humanloop raises $2.6M seed led by Index to help humans teach AI algorithms
Peech raises $8.3M seed led by Ibex for its AI-powered video editing solution
Rill raises $12M seed led by True Ventures for its BI dashboards with embedded database and instant UX
Vetted raises $14M Series A led by Insight for its AI-powered product search engine
Marqvision raises $20M Series A led by DST and Atinum for its AI-powered IP protection platform
Arena raises $32M Series A led by Initialized and Goldcrest Capital to build autopilot for high-frequency decisions
Aisera raises $90M Series D led by Goldman Sachs and Thoma Bravo for its AI-driven service experience platform
If you are getting value from this newsletter, consider subscribing and sharing with your friends: