Last Week in AI

AI for Omicron, Self-Farming Farms, AI-designed beer, Cuddly AI

Our 83rd episode with a summary and discussion of last week's big AI news!
Outline:
(00:00) Intro
(01:10) This AI Software Nearly Predicted Omicron’s Tricky Structure 
(02:48) BioNTech and London A.I. company create "early warning system" for COVID-19 variants
(05:20) We’re one step closer to self-farming farms
(09:55) ‘We Were Blown Away’: How New A.I. Research Is Changing the Way Conservators and Collectors Think About Attribution
(12:42) Meta claims its AI improves speech recognition quality by reading lips
(15:46) The Messy History of Facial Recognition Company Clearview AI
(18:30) Feds' spending on facial recognition tech expands, despite privacy concerns
(20:35) Great Resignation Versus Increasing Investment In AI, Robotics And Automation: A Troubling Trend
(25:55) Github Copilot Wants to Play Chess Instead of Code
(28:22) AIML machine learning students develop beer using AI
(31:40) Robo-dogs and therapy bots: Artificial intelligence goes cuddly
(34:55) Outro
Subscribe: RSS | iTunes | Spotify | YouTube
Feel free to email us your thoughts or feedback at contact@lastweekin.ai

Looking Back at AI in 2021 with Jeremie from Towards Data Science

For our first episode in 2022, we are joined with our friends from the Towards Data Science podcast to discuss our thoughts about the AI-related trends and events that happened in 2021.
Some things we discuss are:
Foundation models continue to grow, but one interesting trend is the focus on efficiency along with (instead of?) scale. For example, while DeepMind’s Gopher model has fewer than twice the parameters of GPT-3, it’s reportedly 25 times more efficient, meaning that much more value is being squeezed out of the same training data and compute. AI21Labs’ Jurrassic models are also equal to GPT-3 on a parameter count basis, but reflect a focus on architecture optimization over raw scaling that we expect to persist into 2022. (That’s not to say significant scaling won’t happen, or that it hasn’t happened already; Microsoft Turing-NLG, released a few months ago, is over half a trillion parameters in size. But it’s safe to say that scaling won’t be done without simultaneous efficiency optimizations that were less of a focus in late-2020.)
Procedural environment generation has been a big theme in reinforcement learning. In Open-Ended Learning Leads to Generally Capable Agents, the team at DeepMind showed how training RL agents on a wide range of environments can lead to emergent behaviour associated with generalization, like trial and error and cooperation with friendly agents.
Open-ended learning (OEL) seems like an interesting wildcard, which some researchers think might be an important ingredient in the final AGI recipe. We spoke with OpenAI’s head of open-ended learning, Ken Stanley, about what role OEL might play in the future of AI on this episode of the TDS podcast.
A NeurIPS spotlight paper titled Optimal Policies Tend to Seek Power, and subsequent work by the same author, are showing that we should expect highly capable AI systems to engage in dangerous behaviour that’s misaligned with human values, by default. Specifically, highly competent agents will tend to search for states that are powerful, in the sense that they offer many downstream options. This finding makes a compelling case that AI alignment ought to be prioritized, particularly given the rate of progress we’re seeing in AI capabilities more broadly. If it really is the case that capable AI systems will be dangerous by default, active effort must be invested in safety research.
Outline:
0:00 Intro
2:15 Rise of multi-modal models
7:40 Growth of hardware and compute
13:20 Reinforcement learning
20:45 Open-ended learning
26:15 Power seeking paper
32:30 Safety and assumptions
35:20 Intrinsic vs. extrinsic motivation
42:00 Mapping natural language
46:20 Timnit Gebru’s research institute
49:20 Wrap-up

Create AI art with an app, an AI-powered game platform, the year of monster AI models, better images of AI

Happy new year!
Our last episode of 2021 is also our 82th episode with a summary and discussion of last week's big AI news!
Outline:
Intro
This AI-powered art app lets you paint pictures with words
AI Dungeon’s creators are launching an experimental AI-powered game platform
Announcing the Transactions on Machine Learning Research
Bigger’s Not Always Better: DeepMind’s New Language AI Is Small But Mighty 
2021 was the year of monster AI models 
U.N. talks adjourn without deal to regulate ‘killer robots’
Face Recognition Is Being Banned—but It’s Still Everywhere
Better Images of AI
Subscribe: RSS | iTunes | Spotify | YouTube
Feel free to email us your thoughts or feedback at contact@lastweekinai.com

AI Best Friends, The Beatles + Machine Learning, Crime Prediction Bias, Transformer Quadraped Robot

Our 81th episode with a summary and discussion of last week's big AI news!
Outline:
(0:00) Intro
(1:30) Meet your new A.I. best friend
(6:37) The Beatles: Get Back Used High-Tech Machine Learning To Restore The Audio
(10:26) Google, Cambridge U & Alan Turing Institute Propose PolyViT: A Universal Transformer for Image, Video, and Audio Classification
(14:00) Deep-learning model speeds extreme weather predictions
(18:23) How We Determined Crime Prediction Software Disproportionately Targeted Low-Income, Black, and Latino Neighborhoods
(23:44) Inside Tesla as Elon Musk Pushed an Unflinching Vision for Self-Driving Cars 
(27:18) Alarmed by Tesla’s public self-driving test, state legislators demand answers from DMV 
(30:15) A ​Quadruped Humanoid Robot Might Be Able To Do It All
(34:05) Forging New Pathways: Boys & Girls Clubs Teens Take AI From Idea to Application
(38:00) Outro
Subscribe: RSS | iTunes | Spotify | YouTube
Feel free to email us your thoughts or feedback at contact@lastweekinai.com

AI for Mathematicians, Timnit Gebru‘s New Research Center, No Ban on Killer Robots, Vertigo AI

Our 80th episode with a summary and discussion of last week's big AI news!
Feel free to email us your thoughts or feedback at contact@lastweekinai.com
Subscribe: RSS | iTunes | Spotify | YouTube
Check our text version of this news roundup over at lastweekin.ai.
Outline:
(01:10) Procedural storytelling is exploding the possibilities of video game narratives (04:16) Twitch Introduces Machine Learning Feature to Detect Suspicious Users(07:28) DeepMind’s AI helps untangle the mathematics of knots(11:06) Yale researchers combat biases in machine learning algorithms(13:40) Ex-Googler Timnit Gebru Starts Her Own AI Research Center(17:20) US rejects calls for regulating or banning ‘killer robots(19:50) Who, exactly, authored this AI-generated spin on Alfred Hitchcock’s Vertigo?  (25:00) Outro
Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

Robots at Google Offices, AI-Powered Writing, AI Polygraph, AI Art

Our 79th episode with a summary and discussion of last week's big AI news!
Feel free to email us your thoughts or feedback at contact@lastweekinai.com
Subscribe: RSS | iTunes | Spotify | YouTube
Check our text version of this news roundup over at lastweekin.ai.
Outline:
(00:00) Intro(01:10) Alphabet is putting its prototype robots to work cleaning up around Google’s offices(04:42) Grammarly raises $200M at a $13B valuation to make you an even better writer using AI  / AI-powered writing assistant Writer nabs $21M(07:03) A Leap Forward in Computer Vision: Facebook AI Says Masked Autoencoders Are Scalable Vision Learners(09:27) Research Professor Adji Bousso Dieng Is Out to Change the World With AI(11:42) A Utah company says it’s revolutionized truth-telling technology. Experts are highly skeptical(13:45) South Korea is Selling Millions of Photos To Facial Recognition Researchers(16:00) AWS makes AI and machine learning tangible with first major art debut at Smithsonian(19:00) ‘Paint Me a Picture’: NVIDIA Research Shows GauGAN AI Art Demo Now Responds to Words(21:45) OutroMusic: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

Chat about Teaching AI at Stanford with Abigail See

A special discussion episode, where we do not cover AI news but instead chat about our experiences teaching AI classes at Stanford with Abigail See, a graduate from the Stanford AI Lab and a good friend of ours.
Subscribe: RSS | iTunes | Spotify | YouTube
Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

Record High AI Funding, AfriBERTa, Facial Recognition in the West Bank, Tesla FSD Beta Crash, AI for Stressed Dogs

Our 78th episode with a summary and discussion of last week's big AI news!
Subscribe: RSS | iTunes | Spotify | YouTube
Check our text version of this news roundup over at lastweekin.ai.
This week:
(00:00) Intro(01:40) Walmart is using fully driverless trucks to ramp up its online grocery business(05:20) AI startup funding hits record high of $17.9B in Q3 / Robot orders by companies surge as labor shortages linger(09:50) University of Waterloo AI Researchers Introduce A New NLP Model ‘AfriBERTa’ For African Languages Using Deep Learning Techniques  (12:40) AI skin cancer diagnoses risk being less accurate for dark skin – study(16:55) Israel escalates surveillance of Palestinians with facial recognition program in West Bank(22:15) Tesla vehicle in ‘Full Self-Driving’ beta mode ‘severely damaged’ after crash in California(26:15) CMU opens first AI maker space to let students 'sharpen the cutting edge of AI'(28:25) Machine Learning Shushes Stressed Dogs(31:05) Outro
 
Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

Facebook Shuts Down Facial Recognition, Google Wants to Work with the Pentagon, AI for Lie Detection?

Our 77th episode with a summary and discussion of last week's big AI news!
Subscribe: RSS | iTunes | Spotify | YouTube
Check our text version of this news roundup over at lastweekin.ai.
This week:
(01:06) Landing AI brings in $57M for its machine learning operations tools
(05:13) As the Arctic Warms, AI Forecasts Scope Out Shifting Sea Ice
(09:07) Why Facebook (Or Meta) Is Making Tactile Sensors for Robots 
(14:12) Google AI Introduces ‘GoEmotions’: An NLP Dataset for Fine-Grained Emotion Classification
(18:28) Facebook, Citing Societal Concerns, Plans to Shut Down Facial Recognition System
(21:40) Google Wants to Work With the Pentagon Again, Despite Employee Concerns
(25:50) Australia Ordered Clearview AI to Destroy its Database, As Its Violating Privacy Laws 
(27:54) AIBA uses AI technology to vet judges and refs at Belgrade worlds
(30:40) Miso Introduces Second Generation Restaurant Kitchen Robot, the Flippy 2
Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)

AI for Game Mods and Moderation, Tesla‘s Beta Rollback, Clearview AI Tested, ruDALL-E

Our 76th episode with a summary and discussion of last week's big AI news!
Subscribe: RSS | iTunes | Spotify | YouTube
Check our text version of this news roundup over at lastweekin.ai.
This week:
AI voice-acting tool xVASynth gets even better at recreating videogame voices 
Microsoft acquires AI-powered moderation platform Two Hat
MIT’s Latest AI Research Using Deep Neural Networks Explains How The Brain Processes Language Works
Making machine learning more useful to high-stakes decision makers
Clearview AI finally takes part in a federal accuracy test.
Tesla pulled the latest FSD Beta from owners’ cars today 
ruDALL-E: generating images from text descriptions, or the largest computational project in Russia
Music: Deliberate Thought, Inspired by Kevin MacLeod (incompetech.com)